├── .gitignore
├── README.md
├── README
└── Logo.png
├── license
├── CHOSUNTRUCK_LICENSE
├── PYUSERINPUT_LICENSE.txt
├── TENSORBOX_LICENSE.txt
└── TENSORFLOW_LICENSE.txt
├── linux
├── Makefile
├── src
│ ├── IPM.cc
│ ├── IPM.h
│ ├── ets2_self_driving.h
│ ├── getScreen_linux.cc
│ ├── getScreen_linux.h
│ ├── linefinder.cc
│ ├── main2.cc
│ ├── uinput.c
│ └── uinput.h
└── tensorbox
│ ├── Car_detection.py
│ ├── data
│ └── inception_v1.ckpt
│ ├── license
│ ├── TENSORBOX_LICENSE.txt
│ └── TENSORFLOW_LICENSE.txt
│ ├── output
│ └── overfeat_rezoom_2017_02_09_13.28
│ │ ├── checkpoint
│ │ ├── hypes.json
│ │ ├── rpc-save.ckpt-100000.val_overlap0.5.txt
│ │ ├── save.ckpt-100000.data-00000-of-00001
│ │ ├── save.ckpt-100000.gt_val.json
│ │ ├── save.ckpt-100000.index
│ │ └── save.ckpt-100000.val.json
│ ├── pymouse
│ ├── __init__.py
│ ├── base.py
│ ├── java_.py
│ ├── mac.py
│ ├── mir.py
│ ├── wayland.py
│ ├── windows.py
│ └── x11.py
│ ├── train.py
│ └── utils
│ ├── Makefile
│ ├── __init__.py
│ ├── annolist
│ ├── AnnoList_pb2.py
│ ├── AnnotationLib.py
│ ├── LICENSE_FOR_THIS_FOLDER
│ ├── MatPlotter.py
│ ├── PalLib.py
│ ├── __init__.py
│ ├── doRPC.py
│ ├── ma_utils.py
│ └── plotSimple.py
│ ├── data_utils.py
│ ├── googlenet_load.py
│ ├── hungarian
│ ├── hungarian.cc
│ ├── hungarian.cpp
│ └── hungarian.hpp
│ ├── rect.py
│ ├── slim_nets
│ ├── __init__.py
│ ├── inception_v1.py
│ ├── resnet_utils.py
│ └── resnet_v1.py
│ ├── stitch_rects.cpp
│ ├── stitch_rects.hpp
│ ├── stitch_wrapper.py
│ ├── stitch_wrapper.pyx
│ └── train_utils.py
└── windows
└── src
├── IPM.cpp
├── IPM.h
├── ets2_self_driving.h
├── guassian_filter.cpp
├── hwnd2mat.cpp
├── linefinder.cpp
└── main.cpp
/.gitignore:
--------------------------------------------------------------------------------
1 | # Created by https://www.gitignore.io/api/c,c++,cuda,opencv,visualstudio,windows
2 |
3 | ### C ###
4 | # Object files
5 | *.o
6 | *.ko
7 | *.obj
8 | *.elf
9 |
10 | # Precompiled Headers
11 | *.gch
12 | *.pch
13 |
14 | # Libraries
15 | *.lib
16 | *.a
17 | *.la
18 | *.lo
19 |
20 | # Shared objects (inc. Windows DLLs)
21 | *.dll
22 | *.so
23 | *.so.*
24 | *.dylib
25 |
26 | # Executables
27 | *.exe
28 | *.out
29 | *.app
30 | *.i*86
31 | *.x86_64
32 | *.hex
33 |
34 | # Debug files
35 | *.dSYM/
36 | *.su
37 |
38 |
39 | ### C++ ###
40 | # Compiled Object files
41 | *.slo
42 | *.lo
43 | *.o
44 | *.obj
45 |
46 | # Precompiled Headers
47 | *.gch
48 | *.pch
49 |
50 | # Compiled Dynamic libraries
51 | *.so
52 | *.dylib
53 | *.dll
54 |
55 | # Fortran module files
56 | *.mod
57 | *.smod
58 |
59 | # Compiled Static libraries
60 | *.lai
61 | *.la
62 | *.a
63 | *.lib
64 |
65 | # Executables
66 | *.exe
67 | *.out
68 | *.app
69 |
70 |
71 | ### CUDA ###
72 | *.i
73 | *.ii
74 | *.gpu
75 | *.ptx
76 | *.cubin
77 | *.fatbin
78 |
79 |
80 | ### OpenCV ###
81 | #OpenCV for Mac and Linux
82 | #build and release folders
83 | */CMakeFiles
84 | */CMakeCache.txt
85 | */cmake_install.cmake
86 | .DS_Store
87 |
88 |
89 | ### VisualStudio ###
90 | ## Ignore Visual Studio temporary files, build results, and
91 | ## files generated by popular Visual Studio add-ons.
92 |
93 | # User-specific files
94 | *.suo
95 | *.user
96 | *.userosscache
97 | *.sln.docstates
98 |
99 | # User-specific files (MonoDevelop/Xamarin Studio)
100 | *.userprefs
101 |
102 | # Build results
103 | [Dd]ebug/
104 | [Dd]ebugPublic/
105 | [Rr]elease/
106 | [Rr]eleases/
107 | x64/
108 | x86/
109 | bld/
110 | [Bb]in/
111 | [Oo]bj/
112 | [Ll]og/
113 |
114 | # Visual Studio 2015 cache/options directory
115 | .vs/
116 | # Uncomment if you have tasks that create the project's static files in wwwroot
117 | #wwwroot/
118 |
119 | # MSTest test Results
120 | [Tt]est[Rr]esult*/
121 | [Bb]uild[Ll]og.*
122 |
123 | # NUNIT
124 | *.VisualState.xml
125 | TestResult.xml
126 |
127 | # Build Results of an ATL Project
128 | [Dd]ebugPS/
129 | [Rr]eleasePS/
130 | dlldata.c
131 |
132 | # DNX
133 | project.lock.json
134 | project.fragment.lock.json
135 | artifacts/
136 |
137 | *_i.c
138 | *_p.c
139 | *_i.h
140 | *.ilk
141 | *.meta
142 | *.obj
143 | *.pch
144 | *.pdb
145 | *.pgc
146 | *.pgd
147 | *.rsp
148 | *.sbr
149 | *.tlb
150 | *.tli
151 | *.tlh
152 | *.tmp
153 | *.tmp_proj
154 | *.log
155 | *.vspscc
156 | *.vssscc
157 | .builds
158 | *.pidb
159 | *.svclog
160 | *.scc
161 |
162 | # Chutzpah Test files
163 | _Chutzpah*
164 |
165 | # Visual C++ cache files
166 | ipch/
167 | *.aps
168 | *.ncb
169 | *.opendb
170 | *.opensdf
171 | *.sdf
172 | *.cachefile
173 | *.VC.db
174 | *.VC.VC.opendb
175 |
176 | # Visual Studio profiler
177 | *.psess
178 | *.vsp
179 | *.vspx
180 | *.sap
181 |
182 | # TFS 2012 Local Workspace
183 | $tf/
184 |
185 | # Guidance Automation Toolkit
186 | *.gpState
187 |
188 | # ReSharper is a .NET coding add-in
189 | _ReSharper*/
190 | *.[Rr]e[Ss]harper
191 | *.DotSettings.user
192 |
193 | # JustCode is a .NET coding add-in
194 | .JustCode
195 |
196 | # TeamCity is a build add-in
197 | _TeamCity*
198 |
199 | # DotCover is a Code Coverage Tool
200 | *.dotCover
201 |
202 | # NCrunch
203 | _NCrunch_*
204 | .*crunch*.local.xml
205 | nCrunchTemp_*
206 |
207 | # MightyMoose
208 | *.mm.*
209 | AutoTest.Net/
210 |
211 | # Web workbench (sass)
212 | .sass-cache/
213 |
214 | # Installshield output folder
215 | [Ee]xpress/
216 |
217 | # DocProject is a documentation generator add-in
218 | DocProject/buildhelp/
219 | DocProject/Help/*.HxT
220 | DocProject/Help/*.HxC
221 | DocProject/Help/*.hhc
222 | DocProject/Help/*.hhk
223 | DocProject/Help/*.hhp
224 | DocProject/Help/Html2
225 | DocProject/Help/html
226 |
227 | # Click-Once directory
228 | publish/
229 |
230 | # Publish Web Output
231 | *.[Pp]ublish.xml
232 | *.azurePubxml
233 | # TODO: Comment the next line if you want to checkin your web deploy settings
234 | # but database connection strings (with potential passwords) will be unencrypted
235 | *.pubxml
236 | *.publishproj
237 |
238 | # Microsoft Azure Web App publish settings. Comment the next line if you want to
239 | # checkin your Azure Web App publish settings, but sensitive information contained
240 | # in these scripts will be unencrypted
241 | PublishScripts/
242 |
243 | # NuGet Packages
244 | *.nupkg
245 | # The packages folder can be ignored because of Package Restore
246 | **/packages/*
247 | # except build/, which is used as an MSBuild target.
248 | !**/packages/build/
249 | # Uncomment if necessary however generally it will be regenerated when needed
250 | #!**/packages/repositories.config
251 | # NuGet v3's project.json files produces more ignoreable files
252 | *.nuget.props
253 | *.nuget.targets
254 |
255 | # Microsoft Azure Build Output
256 | csx/
257 | *.build.csdef
258 |
259 | # Microsoft Azure Emulator
260 | ecf/
261 | rcf/
262 |
263 | # Windows Store app package directories and files
264 | AppPackages/
265 | BundleArtifacts/
266 | Package.StoreAssociation.xml
267 | _pkginfo.txt
268 |
269 | # Visual Studio cache files
270 | # files ending in .cache can be ignored
271 | *.[Cc]ache
272 | # but keep track of directories ending in .cache
273 | !*.[Cc]ache/
274 |
275 | # Others
276 | ClientBin/
277 | ~$*
278 | *~
279 | *.dbmdl
280 | *.dbproj.schemaview
281 | *.pfx
282 | *.publishsettings
283 | node_modules/
284 | orleans.codegen.cs
285 |
286 | # Since there are multiple workflows, uncomment next line to ignore bower_components
287 | # (https://github.com/github/gitignore/pull/1529#issuecomment-104372622)
288 | #bower_components/
289 |
290 | # RIA/Silverlight projects
291 | Generated_Code/
292 |
293 | # Backup & report files from converting an old project file
294 | # to a newer Visual Studio version. Backup files are not needed,
295 | # because we have git ;-)
296 | _UpgradeReport_Files/
297 | Backup*/
298 | UpgradeLog*.XML
299 | UpgradeLog*.htm
300 |
301 | # SQL Server files
302 | *.mdf
303 | *.ldf
304 |
305 | # Business Intelligence projects
306 | *.rdl.data
307 | *.bim.layout
308 | *.bim_*.settings
309 |
310 | # Microsoft Fakes
311 | FakesAssemblies/
312 |
313 | # GhostDoc plugin setting file
314 | *.GhostDoc.xml
315 |
316 | # Node.js Tools for Visual Studio
317 | .ntvs_analysis.dat
318 |
319 | # Visual Studio 6 build log
320 | *.plg
321 |
322 | # Visual Studio 6 workspace options file
323 | *.opt
324 |
325 | # Visual Studio LightSwitch build output
326 | **/*.HTMLClient/GeneratedArtifacts
327 | **/*.DesktopClient/GeneratedArtifacts
328 | **/*.DesktopClient/ModelManifest.xml
329 | **/*.Server/GeneratedArtifacts
330 | **/*.Server/ModelManifest.xml
331 | _Pvt_Extensions
332 |
333 | # Paket dependency manager
334 | .paket/paket.exe
335 | paket-files/
336 |
337 | # FAKE - F# Make
338 | .fake/
339 |
340 | # JetBrains Rider
341 | .idea/
342 | *.sln.iml
343 |
344 |
345 | ### Windows ###
346 | # Windows image file caches
347 | Thumbs.db
348 | ehthumbs.db
349 |
350 | # Folder config file
351 | Desktop.ini
352 |
353 | # Recycle Bin used on file shares
354 | $RECYCLE.BIN/
355 |
356 | # Windows Installer files
357 | *.cab
358 | *.msi
359 | *.msm
360 | *.msp
361 |
362 | # Windows shortcuts
363 | *.lnk
364 |
365 | # Generated files
366 | ChosunTruck
367 |
368 | # Build files
369 | build/
370 |
371 | # pyc files
372 | *.pyc
373 |
--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | #
ChosunTruck
2 |
3 | ## Introduction
4 | ChosunTruck is an autonomous driving solution for [Euro Truck Simulator 2](https://eurotrucksimulator2.com/).
5 | Recently, autonomous driving technology has become a big issue and as a result we have been studying technology that incorporates this.
6 | It is being developed in a simulated environment called Euro Truck Simulator 2 to allow us to study it using vehicles.
7 | We chose Euro Truck Simulator 2 because this simulator provides a good test environment that is similar to the real road.
8 |
9 | ## Features
10 | * You can drive a vehicle without handling it yourself.
11 | * You can understand the principles of autonomous driving.
12 | * (Experimental/Linux only) You can detect where other vehicles are.
13 |
14 | ## How To Run It
15 | ### Windows
16 |
17 | #### Dependencies
18 | - OS: Windows 7, 10 (64bit)
19 |
20 | - IDE: Visual Studio 2013, 2015
21 |
22 | - OpenCV version: >= 3.1
23 |
24 | - [Cuda Toolkit 7.5](https://developer.nvidia.com/cuda-75-downloads-archive) (Note: Do an ADVANCED INSTALLATION. ONLY install the Toolkit + Integration to Visual Studio. Do NOT install the drivers + other stuff it would normally give you. Once installed, your project properties should look like this: https://i.imgur.com/e7IRtjy.png)
25 |
26 | - If you have a problem during installation, look at our [Windows Installation wiki page](https://github.com/bethesirius/ChosunTruck/wiki/Windows-Installation)
27 |
28 | #### Required to allow input to work in Windows:
29 | - **Go to C:\Users\YOURUSERNAME\Documents\Euro Truck Simulator 2\profiles and edit controls.sii from**
30 | ```
31 | config_lines[0]: "device keyboard `di8.keyboard`"
32 | config_lines[1]: "device mouse `fusion.mouse`"
33 | ```
34 | to
35 | ```
36 | config_lines[0]: "device keyboard `sys.keyboard`"
37 | config_lines[1]: "device mouse `sys.mouse`"
38 | ```
39 | (thanks Komat!)
40 | - **While you are in controls.sii, make sure your sensitivity is set to:**
41 | ```
42 | config_lines[33]: "constant c_rsteersens 0.775000"
43 | config_lines[34]: "constant c_asteersens 4.650000"
44 | ```
45 | #### Then:
46 | - Set controls.sii to read-only
47 | - Open the visual studio project and build it.
48 | - Run ETS2 in windowed mode and set resolution to 1024 * 768.(It will work properly with 1920 * 1080 screen resolution and 1024 * 768 window mode ETS2.)
49 |
50 | ### Linux
51 | #### Dependencies
52 | - OS: Ubuntu 16.04 LTS
53 |
54 | - [OpenCV version: >= 3.1](http://embedonix.com/articles/image-processing/installing-opencv-3-1-0-on-ubuntu/)
55 |
56 | - (Optional) Tensorflow version: >= 0.12.1
57 |
58 | ### Build the source code with the following command (inside the linux directory).
59 | ```
60 | make
61 | ```
62 | ### If you want the car detection function then:
63 | ````
64 | make Drive
65 | ````
66 | #### Then:
67 | - Run ETS2 in windowed mode and set its resolution to 1024 * 768. (It will work properly with 1920 * 1080 screen resolution and 1024 * 768 windowed mode ETS2)
68 | - It cannot find the ETS2 window automatically. Move the ETS2 window to the right-down corner to fix this.
69 | - In ETS2 Options, set controls to 'Keyboard + Mouse Steering', 'left click' to acclerate, and 'right click' to brake.
70 | - Go to a highway and set the truck's speed to 40~60km/h. (I recommend you turn on cruise mode to set the speed easily)
71 | - Run this program!
72 |
73 | #### To enable car detection mode, add -D or --Car_Detection.
74 | ```
75 | ./ChosunTruck [-D|--Car_Detection]
76 | ```
77 | ## Troubleshooting
78 | See [Our wiki page](https://github.com/bethesirius/ChosunTruck/wiki/Troubleshooting).
79 |
80 | If you have some problems running this project, reference the demo video below. Or, [open a issue to contact our team](https://github.com/bethesirius/ChosunTruck/issues).
81 |
82 | ## Demo Video
83 | Lane Detection (Youtube link)
84 |
85 | [](http://www.youtube.com/watch?v=vF7J_uC045Q)
86 | [](http://www.youtube.com/watch?v=qb99czlIklA)
87 |
88 | Lane Detection + Vehicle Detection (Youtube link)
89 |
90 | [](http://www.youtube.com/watch?v=w6H2eGEvzvw)
91 |
92 | ## Todo
93 | * For better detection performance, Change the Tensorbox to YOLO2.
94 | * The information from in-game screen have Restrictions. Read ETS2 process memory to collect more driving environment data.
95 |
96 | ## Founders
97 | - Chiwan Song, chi3236@gmail.com
98 |
99 | - JaeCheol Sim, simjaecheol@naver.com
100 |
101 | - Seongjoon Chu, hs4393@gmail.com
102 |
103 | ## Contributors
104 | - [zappybiby](https://github.com/zappybiby)
105 |
106 | ## How To Contribute
107 | Anyone who is interested in this project is welcome! Just fork it and pull requests!
108 |
109 | ## License
110 | ChosunTruck, Euro Truck Simulator 2 auto driving solution
111 | Copyright (C) 2017 chi3236, bethesirius, uoyssim
112 |
113 | This program is free software: you can redistribute it and/or modify
114 | it under the terms of the GNU General Public License as published by
115 | the Free Software Foundation, either version 3 of the License, or
116 | (at your option) any later version.
117 |
118 | This program is distributed in the hope that it will be useful,
119 | but WITHOUT ANY WARRANTY; without even the implied warranty of
120 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
121 | GNU General Public License for more details.
122 |
123 | You should have received a copy of the GNU General Public License
124 | along with this program. If not, see .
125 |
--------------------------------------------------------------------------------
/README/Logo.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/bethesirius/ChosunTruck/889644385ce57f971ec2921f006fbb0a167e6f1e/README/Logo.png
--------------------------------------------------------------------------------
/license/TENSORBOX_LICENSE.txt:
--------------------------------------------------------------------------------
1 | Copyright (c) 2016 "The Contributors"
2 |
3 |
4 | Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
5 |
6 | The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
7 |
8 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
9 |
--------------------------------------------------------------------------------
/license/TENSORFLOW_LICENSE.txt:
--------------------------------------------------------------------------------
1 | Copyright 2015 The TensorFlow Authors. All rights reserved.
2 |
3 | Apache License
4 | Version 2.0, January 2004
5 | http://www.apache.org/licenses/
6 |
7 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
8 |
9 | 1. Definitions.
10 |
11 | "License" shall mean the terms and conditions for use, reproduction,
12 | and distribution as defined by Sections 1 through 9 of this document.
13 |
14 | "Licensor" shall mean the copyright owner or entity authorized by
15 | the copyright owner that is granting the License.
16 |
17 | "Legal Entity" shall mean the union of the acting entity and all
18 | other entities that control, are controlled by, or are under common
19 | control with that entity. For the purposes of this definition,
20 | "control" means (i) the power, direct or indirect, to cause the
21 | direction or management of such entity, whether by contract or
22 | otherwise, or (ii) ownership of fifty percent (50%) or more of the
23 | outstanding shares, or (iii) beneficial ownership of such entity.
24 |
25 | "You" (or "Your") shall mean an individual or Legal Entity
26 | exercising permissions granted by this License.
27 |
28 | "Source" form shall mean the preferred form for making modifications,
29 | including but not limited to software source code, documentation
30 | source, and configuration files.
31 |
32 | "Object" form shall mean any form resulting from mechanical
33 | transformation or translation of a Source form, including but
34 | not limited to compiled object code, generated documentation,
35 | and conversions to other media types.
36 |
37 | "Work" shall mean the work of authorship, whether in Source or
38 | Object form, made available under the License, as indicated by a
39 | copyright notice that is included in or attached to the work
40 | (an example is provided in the Appendix below).
41 |
42 | "Derivative Works" shall mean any work, whether in Source or Object
43 | form, that is based on (or derived from) the Work and for which the
44 | editorial revisions, annotations, elaborations, or other modifications
45 | represent, as a whole, an original work of authorship. For the purposes
46 | of this License, Derivative Works shall not include works that remain
47 | separable from, or merely link (or bind by name) to the interfaces of,
48 | the Work and Derivative Works thereof.
49 |
50 | "Contribution" shall mean any work of authorship, including
51 | the original version of the Work and any modifications or additions
52 | to that Work or Derivative Works thereof, that is intentionally
53 | submitted to Licensor for inclusion in the Work by the copyright owner
54 | or by an individual or Legal Entity authorized to submit on behalf of
55 | the copyright owner. For the purposes of this definition, "submitted"
56 | means any form of electronic, verbal, or written communication sent
57 | to the Licensor or its representatives, including but not limited to
58 | communication on electronic mailing lists, source code control systems,
59 | and issue tracking systems that are managed by, or on behalf of, the
60 | Licensor for the purpose of discussing and improving the Work, but
61 | excluding communication that is conspicuously marked or otherwise
62 | designated in writing by the copyright owner as "Not a Contribution."
63 |
64 | "Contributor" shall mean Licensor and any individual or Legal Entity
65 | on behalf of whom a Contribution has been received by Licensor and
66 | subsequently incorporated within the Work.
67 |
68 | 2. Grant of Copyright License. Subject to the terms and conditions of
69 | this License, each Contributor hereby grants to You a perpetual,
70 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable
71 | copyright license to reproduce, prepare Derivative Works of,
72 | publicly display, publicly perform, sublicense, and distribute the
73 | Work and such Derivative Works in Source or Object form.
74 |
75 | 3. Grant of Patent License. Subject to the terms and conditions of
76 | this License, each Contributor hereby grants to You a perpetual,
77 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable
78 | (except as stated in this section) patent license to make, have made,
79 | use, offer to sell, sell, import, and otherwise transfer the Work,
80 | where such license applies only to those patent claims licensable
81 | by such Contributor that are necessarily infringed by their
82 | Contribution(s) alone or by combination of their Contribution(s)
83 | with the Work to which such Contribution(s) was submitted. If You
84 | institute patent litigation against any entity (including a
85 | cross-claim or counterclaim in a lawsuit) alleging that the Work
86 | or a Contribution incorporated within the Work constitutes direct
87 | or contributory patent infringement, then any patent licenses
88 | granted to You under this License for that Work shall terminate
89 | as of the date such litigation is filed.
90 |
91 | 4. Redistribution. You may reproduce and distribute copies of the
92 | Work or Derivative Works thereof in any medium, with or without
93 | modifications, and in Source or Object form, provided that You
94 | meet the following conditions:
95 |
96 | (a) You must give any other recipients of the Work or
97 | Derivative Works a copy of this License; and
98 |
99 | (b) You must cause any modified files to carry prominent notices
100 | stating that You changed the files; and
101 |
102 | (c) You must retain, in the Source form of any Derivative Works
103 | that You distribute, all copyright, patent, trademark, and
104 | attribution notices from the Source form of the Work,
105 | excluding those notices that do not pertain to any part of
106 | the Derivative Works; and
107 |
108 | (d) If the Work includes a "NOTICE" text file as part of its
109 | distribution, then any Derivative Works that You distribute must
110 | include a readable copy of the attribution notices contained
111 | within such NOTICE file, excluding those notices that do not
112 | pertain to any part of the Derivative Works, in at least one
113 | of the following places: within a NOTICE text file distributed
114 | as part of the Derivative Works; within the Source form or
115 | documentation, if provided along with the Derivative Works; or,
116 | within a display generated by the Derivative Works, if and
117 | wherever such third-party notices normally appear. The contents
118 | of the NOTICE file are for informational purposes only and
119 | do not modify the License. You may add Your own attribution
120 | notices within Derivative Works that You distribute, alongside
121 | or as an addendum to the NOTICE text from the Work, provided
122 | that such additional attribution notices cannot be construed
123 | as modifying the License.
124 |
125 | You may add Your own copyright statement to Your modifications and
126 | may provide additional or different license terms and conditions
127 | for use, reproduction, or distribution of Your modifications, or
128 | for any such Derivative Works as a whole, provided Your use,
129 | reproduction, and distribution of the Work otherwise complies with
130 | the conditions stated in this License.
131 |
132 | 5. Submission of Contributions. Unless You explicitly state otherwise,
133 | any Contribution intentionally submitted for inclusion in the Work
134 | by You to the Licensor shall be under the terms and conditions of
135 | this License, without any additional terms or conditions.
136 | Notwithstanding the above, nothing herein shall supersede or modify
137 | the terms of any separate license agreement you may have executed
138 | with Licensor regarding such Contributions.
139 |
140 | 6. Trademarks. This License does not grant permission to use the trade
141 | names, trademarks, service marks, or product names of the Licensor,
142 | except as required for reasonable and customary use in describing the
143 | origin of the Work and reproducing the content of the NOTICE file.
144 |
145 | 7. Disclaimer of Warranty. Unless required by applicable law or
146 | agreed to in writing, Licensor provides the Work (and each
147 | Contributor provides its Contributions) on an "AS IS" BASIS,
148 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
149 | implied, including, without limitation, any warranties or conditions
150 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
151 | PARTICULAR PURPOSE. You are solely responsible for determining the
152 | appropriateness of using or redistributing the Work and assume any
153 | risks associated with Your exercise of permissions under this License.
154 |
155 | 8. Limitation of Liability. In no event and under no legal theory,
156 | whether in tort (including negligence), contract, or otherwise,
157 | unless required by applicable law (such as deliberate and grossly
158 | negligent acts) or agreed to in writing, shall any Contributor be
159 | liable to You for damages, including any direct, indirect, special,
160 | incidental, or consequential damages of any character arising as a
161 | result of this License or out of the use or inability to use the
162 | Work (including but not limited to damages for loss of goodwill,
163 | work stoppage, computer failure or malfunction, or any and all
164 | other commercial damages or losses), even if such Contributor
165 | has been advised of the possibility of such damages.
166 |
167 | 9. Accepting Warranty or Additional Liability. While redistributing
168 | the Work or Derivative Works thereof, You may choose to offer,
169 | and charge a fee for, acceptance of support, warranty, indemnity,
170 | or other liability obligations and/or rights consistent with this
171 | License. However, in accepting such obligations, You may act only
172 | on Your own behalf and on Your sole responsibility, not on behalf
173 | of any other Contributor, and only if You agree to indemnify,
174 | defend, and hold each Contributor harmless for any liability
175 | incurred by, or claims asserted against, such Contributor by reason
176 | of your accepting any such warranty or additional liability.
177 |
178 | END OF TERMS AND CONDITIONS
179 |
180 | APPENDIX: How to apply the Apache License to your work.
181 |
182 | To apply the Apache License to your work, attach the following
183 | boilerplate notice, with the fields enclosed by brackets "[]"
184 | replaced with your own identifying information. (Don't include
185 | the brackets!) The text should be enclosed in the appropriate
186 | comment syntax for the file format. We also recommend that a
187 | file or class name and description of purpose be included on the
188 | same "printed page" as the copyright notice for easier
189 | identification within third-party archives.
190 |
191 | Copyright 2015, The TensorFlow Authors.
192 |
193 | Licensed under the Apache License, Version 2.0 (the "License");
194 | you may not use this file except in compliance with the License.
195 | You may obtain a copy of the License at
196 |
197 | http://www.apache.org/licenses/LICENSE-2.0
198 |
199 | Unless required by applicable law or agreed to in writing, software
200 | distributed under the License is distributed on an "AS IS" BASIS,
201 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
202 | See the License for the specific language governing permissions and
203 | limitations under the License.
204 |
--------------------------------------------------------------------------------
/linux/Makefile:
--------------------------------------------------------------------------------
1 | OUT_O_DIR = build
2 | TENSORBOX_UTILS_DIR = tensorbox/utils
3 |
4 | OBJS = $(OUT_O_DIR)/getScreen_linux.o $(OUT_O_DIR)/main2.o $(OUT_O_DIR)/IPM.o $(OUT_O_DIR)/linefinder.o $(OUT_O_DIR)/uinput.o
5 |
6 | CC = gcc
7 | CFLAGS = -std=c11 -Wall -O3 -march=native
8 | CPP = g++
9 | CPPFLAGS = `pkg-config opencv --cflags --libs` -std=c++11 -lX11 -Wall -fopenmp -O3 -march=native
10 |
11 | TARGET = ChosunTruck
12 |
13 | $(TARGET) : $(OBJS)
14 | $(CPP) $(OBJS) $(CPPFLAGS) -o $@
15 |
16 | $(OUT_O_DIR)/main2.o : src/main2.cc
17 | mkdir -p $(@D)
18 | $(CPP) -c $< $(CPPFLAGS) -o $@
19 | $(OUT_O_DIR)/getScreen_linux.o : src/getScreen_linux.cc
20 | mkdir -p $(@D)
21 | $(CPP) -c $< $(CPPFLAGS) -o $@
22 | $(OUT_O_DIR)/IPM.o : src/IPM.cc
23 | mkdir -p $(@D)
24 | $(CPP) -c $< $(CPPFLAGS) -o $@
25 | $(OUT_O_DIR)/linefinder.o : src/linefinder.cc
26 | mkdir -p $(@D)
27 | $(CPP) -c $< $(CPPFLAGS) -o $@
28 | $(OUT_O_DIR)/uinput.o : src/uinput.c
29 | mkdir -p $(@D)
30 | $(CC) -c $< $(CFLAGS) -o $@
31 |
32 | clean :
33 | rm -f $(OBJS) ./$(TARGET)
34 |
35 | .PHONY: Drive
36 |
37 | Drive:
38 | pip install runcython
39 | makecython++ $(TENSORBOX_UTILS_DIR)/stitch_wrapper.pyx "" "$(TENSORBOX_UTILS_DIR)/stitch_rects.cpp $(TENSORBOX_UTILS_DIR)/hungarian/hungarian.cpp"
40 |
41 | hungarian: $(TENSORBOX_UTILS_DIR)/hungarian/hungarian.so
42 |
43 | $(TENSORBOX_UTILS_DIR)/hungarian/hungarian.so:
44 | cd $(TENSORBOX_UTILS_DIR)/hungarian && \
45 | TF_INC=$$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_include())') && \
46 | if [ `uname` == Darwin ];\
47 | then g++ -std=c++11 -shared hungarian.cc -o hungarian.so -fPIC -I -D_GLIBCXX_USE_CXX11_ABI=0$$TF_INC;\
48 | else g++ -std=c++11 -shared hungarian.cc -o hungarian.so -fPIC -I $$TF_INC; fi
49 |
50 |
51 |
--------------------------------------------------------------------------------
/linux/src/IPM.cc:
--------------------------------------------------------------------------------
1 | #include "IPM.h"
2 |
3 | using namespace cv;
4 | using namespace std;
5 |
6 | // Public
7 | IPM::IPM( const cv::Size& _origSize, const cv::Size& _dstSize, const std::vector& _origPoints, const std::vector& _dstPoints )
8 | : m_origSize(_origSize), m_dstSize(_dstSize), m_origPoints(_origPoints), m_dstPoints(_dstPoints)
9 | {
10 | assert( m_origPoints.size() == 4 && m_dstPoints.size() == 4 && "Orig. points and Dst. points must vectors of 4 points" );
11 | m_H = getPerspectiveTransform( m_origPoints, m_dstPoints );
12 | m_H_inv = m_H.inv();
13 |
14 | createMaps();
15 | }
16 |
17 | void IPM::setIPM( const cv::Size& _origSize, const cv::Size& _dstSize, const std::vector& _origPoints, const std::vector& _dstPoints )
18 | {
19 | m_origSize = _origSize;
20 | m_dstSize = _dstSize;
21 | m_origPoints = _origPoints;
22 | m_dstPoints = _dstPoints;
23 | assert( m_origPoints.size() == 4 && m_dstPoints.size() == 4 && "Orig. points and Dst. points must vectors of 4 points" );
24 | m_H = getPerspectiveTransform( m_origPoints, m_dstPoints );
25 | m_H_inv = m_H.inv();
26 |
27 | createMaps();
28 | }
29 | void IPM::drawPoints( const std::vector& _points, cv::Mat& _img ) const
30 | {
31 | assert(_points.size() == 4);
32 |
33 | line(_img, Point(static_cast(_points[0].x), static_cast(_points[0].y)), Point(static_cast(_points[3].x), static_cast(_points[3].y)), CV_RGB( 205,205,0), 2);
34 | line(_img, Point(static_cast(_points[2].x), static_cast(_points[2].y)), Point(static_cast(_points[3].x), static_cast(_points[3].y)), CV_RGB( 205,205,0), 2);
35 | line(_img, Point(static_cast(_points[0].x), static_cast(_points[0].y)), Point(static_cast(_points[1].x), static_cast(_points[1].y)), CV_RGB( 205,205,0), 2);
36 | line(_img, Point(static_cast(_points[2].x), static_cast(_points[2].y)), Point(static_cast(_points[1].x), static_cast(_points[1].y)), CV_RGB( 205,205,0), 2);
37 | for(size_t i=0; i<_points.size(); i++)
38 | {
39 | circle(_img, Point(static_cast(_points[i].x), static_cast(_points[i].y)), 2, CV_RGB(238,238,0), -1);
40 | circle(_img, Point(static_cast(_points[i].x), static_cast(_points[i].y)), 5, CV_RGB(255,255,255), 2);
41 | }
42 | }
43 | void IPM::getPoints(vector& _origPts, vector& _ipmPts)
44 | {
45 | _origPts = m_origPoints;
46 | _ipmPts = m_dstPoints;
47 | }
48 | void IPM::applyHomography(const Mat& _inputImg, Mat& _dstImg, int _borderMode)
49 | {
50 | // Generate IPM image from src
51 | remap(_inputImg, _dstImg, m_mapX, m_mapY, INTER_LINEAR, _borderMode);//, BORDER_CONSTANT, Scalar(0,0,0,0));
52 | }
53 | void IPM::applyHomographyInv(const Mat& _inputImg, Mat& _dstImg, int _borderMode)
54 | {
55 | // Generate IPM image from src
56 | remap(_inputImg, _dstImg, m_mapX, m_mapY, INTER_LINEAR, _borderMode);//, BORDER_CONSTANT, Scalar(0,0,0,0));
57 | }
58 | Point2d IPM::applyHomography( const Point2d& _point )
59 | {
60 | return applyHomography( _point, m_H );
61 | }
62 | Point2d IPM::applyHomographyInv( const Point2d& _point )
63 | {
64 | return applyHomography( _point, m_H_inv );
65 | }
66 | Point2d IPM::applyHomography( const Point2d& _point, const Mat& _H )
67 | {
68 | Point2d ret = Point2d( -1, -1 );
69 |
70 | const double u = _H.at(0,0) * _point.x + _H.at(0,1) * _point.y + _H.at(0,2);
71 | const double v = _H.at(1,0) * _point.x + _H.at(1,1) * _point.y + _H.at(1,2);
72 | const double s = _H.at(2,0) * _point.x + _H.at(2,1) * _point.y + _H.at(2,2);
73 | if ( s != 0 )
74 | {
75 | ret.x = ( u / s );
76 | ret.y = ( v / s );
77 | }
78 | return ret;
79 | }
80 | Point3d IPM::applyHomography( const Point3d& _point )
81 | {
82 | return applyHomography( _point, m_H );
83 | }
84 | Point3d IPM::applyHomographyInv( const Point3d& _point )
85 | {
86 | return applyHomography( _point, m_H_inv );
87 | }
88 | Point3d IPM::applyHomography( const Point3d& _point, const cv::Mat& _H )
89 | {
90 | Point3d ret = Point3d( -1, -1, 1 );
91 |
92 | const double u = _H.at(0,0) * _point.x + _H.at(0,1) * _point.y + _H.at(0,2) * _point.z;
93 | const double v = _H.at(1,0) * _point.x + _H.at(1,1) * _point.y + _H.at(1,2) * _point.z;
94 | const double s = _H.at(2,0) * _point.x + _H.at(2,1) * _point.y + _H.at(2,2) * _point.z;
95 | if ( s != 0 )
96 | {
97 | ret.x = ( u / s );
98 | ret.y = ( v / s );
99 | }
100 | else
101 | ret.z = 0;
102 | return ret;
103 | }
104 |
105 | // Private
106 | void IPM::createMaps()
107 | {
108 | // Create remap images
109 | m_mapX.create(m_dstSize, CV_32F);
110 | m_mapY.create(m_dstSize, CV_32F);
111 | //#pragma omp parallel for schedule(dynamic)
112 | for( int j = 0; j < m_dstSize.height; ++j )
113 | {
114 | float* ptRowX = m_mapX.ptr(j);
115 | float* ptRowY = m_mapY.ptr(j);
116 | //#pragma omp parallel for schedule(dynamic)
117 | for( int i = 0; i < m_dstSize.width; ++i )
118 | {
119 | Point2f pt = applyHomography( Point2f( static_cast(i), static_cast(j) ), m_H_inv );
120 | ptRowX[i] = pt.x;
121 | ptRowY[i] = pt.y;
122 | }
123 | }
124 |
125 | m_invMapX.create(m_origSize, CV_32F);
126 | m_invMapY.create(m_origSize, CV_32F);
127 |
128 | //#pragma omp parallel for schedule(dynamic)
129 | for( int j = 0; j < m_origSize.height; ++j )
130 | {
131 | float* ptRowX = m_invMapX.ptr(j);
132 | float* ptRowY = m_invMapY.ptr(j);
133 | //#pragma omp parallel for schedule(dynamic)
134 | for( int i = 0; i < m_origSize.width; ++i )
135 | {
136 | Point2f pt = applyHomography( Point2f( static_cast(i), static_cast(j) ), m_H );
137 | ptRowX[i] = pt.x;
138 | ptRowY[i] = pt.y;
139 | }
140 | }
141 | }
142 |
--------------------------------------------------------------------------------
/linux/src/IPM.h:
--------------------------------------------------------------------------------
1 | #ifndef __IPM_H__
2 | #define __IPM_H__
3 |
4 | #include
5 | #include
6 | #include
7 | #include
8 |
9 | #include
10 |
11 | class IPM
12 | {
13 | public:
14 | IPM( const cv::Size& _origSize, const cv::Size& _dstSize,
15 | const std::vector& _origPoints, const std::vector& _dstPoints );
16 |
17 | IPM() {}
18 | // Apply IPM on points
19 | cv::Point2d applyHomography(const cv::Point2d& _point, const cv::Mat& _H);
20 | cv::Point3d applyHomography( const cv::Point3d& _point, const cv::Mat& _H);
21 | cv::Point2d applyHomography(const cv::Point2d& _point);
22 | cv::Point3d applyHomography( const cv::Point3d& _point);
23 | cv::Point2d applyHomographyInv(const cv::Point2d& _point);
24 | cv::Point3d applyHomographyInv( const cv::Point3d& _point);
25 | void applyHomography( const cv::Mat& _origBGR, cv::Mat& _ipmBGR, int borderMode = cv::BORDER_CONSTANT);
26 | void applyHomographyInv( const cv::Mat& _ipmBGR, cv::Mat& _origBGR, int borderMode = cv::BORDER_CONSTANT);
27 |
28 | // Getters
29 | cv::Mat getH() const { return m_H; }
30 | cv::Mat getHinv() const { return m_H_inv; }
31 | void getPoints(std::vector& _origPts, std::vector& _ipmPts);
32 |
33 | // Draw
34 | void drawPoints( const std::vector& _points, cv::Mat& _img ) const;
35 |
36 | void setIPM( const cv::Size& _origSize, const cv::Size& _dstSize,
37 | const std::vector& _origPoints, const std::vector& _dstPoints );
38 | private:
39 | void createMaps();
40 |
41 | // Sizes
42 | cv::Size m_origSize;
43 | cv::Size m_dstSize;
44 |
45 | // Points
46 | std::vector m_origPoints;
47 | std::vector m_dstPoints;
48 |
49 | // Homography
50 | cv::Mat m_H;
51 | cv::Mat m_H_inv;
52 |
53 | // Maps
54 | cv::Mat m_mapX, m_mapY;
55 | cv::Mat m_invMapX, m_invMapY;
56 | };
57 |
58 | #endif /*__IPM_H__*/
59 |
--------------------------------------------------------------------------------
/linux/src/ets2_self_driving.h:
--------------------------------------------------------------------------------
1 | #ifndef __ets_self_driving_h__
2 | #define __ets_self_driving_h__
3 |
4 | #include
5 | #include
6 | //#include
7 | #include
8 | #include
9 | #include
10 | #include
11 | //#include
12 | //#include
13 | //#include
14 | //#include
15 | #include
16 | #include
17 | #include
18 |
19 | #define PI 3.1415926
20 |
21 |
22 | using namespace cv;
23 | using namespace std;
24 |
25 | class LineFinder{
26 |
27 | private:
28 | cv::Mat image; // 원 영상
29 | std::vector lines; // 선을 감지하기 위한 마지막 점을 포함한 벡터
30 | double deltaRho;
31 | double deltaTheta; // 누산기 해상도 파라미터
32 | int minVote; // 선을 고려하기 전에 받아야 하는 최소 투표 개수
33 | double minLength; // 선에 대한 최소 길이
34 | double maxGap; // 선에 따른 최대 허용 간격
35 |
36 | public:
37 | LineFinder() : deltaRho(1), deltaTheta(PI / 180), minVote(50), minLength(50), maxGap(10) {}
38 | // 기본 누적 해상도는 1각도 1화소
39 | // 간격이 없고 최소 길이도 없음
40 | void setAccResolution(double dRho, double dTheta);
41 | void setMinVote(int minv);
42 | void setLineLengthAndGap(double length, double gap);
43 | std::vector findLines(cv::Mat& binary);
44 | void drawDetectedLines(cv::Mat &image, cv::Scalar color = cv::Scalar(112, 112, 0));
45 | };
46 | //Mat hwnd2mat(HWND hwnd);
47 | void cudaf();
48 |
49 | #endif
50 |
--------------------------------------------------------------------------------
/linux/src/getScreen_linux.cc:
--------------------------------------------------------------------------------
1 | #include
2 | #include
3 | #include
4 | #include
5 | #include
6 |
7 | void ImageFromDisplay(std::vector& Pixels, int& Width, int& Height, int& BitsPerPixel)
8 | {
9 | Display* display = XOpenDisplay(nullptr);
10 | Window root = DefaultRootWindow(display);
11 |
12 | XWindowAttributes attributes = {0};
13 | XGetWindowAttributes(display, root, &attributes);
14 |
15 | Width = attributes.width;
16 | Height = attributes.height;
17 |
18 | XImage* img = XGetImage(display, root, 0, 0 , Width, Height, AllPlanes, ZPixmap);
19 | BitsPerPixel = img->bits_per_pixel;
20 | Pixels.resize(Width * Height * 4);
21 |
22 | memcpy(&Pixels[0], img->data, Pixels.size());
23 |
24 | XDestroyImage(img);
25 | XCloseDisplay(display);
26 | }
27 |
--------------------------------------------------------------------------------
/linux/src/getScreen_linux.h:
--------------------------------------------------------------------------------
1 | #include
2 | #include
3 | #include
4 | #include
5 | #include
6 |
7 | void ImageFromDisplay(std::vector& Pixels, int& Width, int& Height, int& BitsPerPixel);
8 |
--------------------------------------------------------------------------------
/linux/src/linefinder.cc:
--------------------------------------------------------------------------------
1 | #include "ets2_self_driving.h"
2 | #include
3 | #include
4 | //#include
5 | #include
6 | #include
7 | #include
8 | #include
9 | //#include
10 | //#include
11 | //#include
12 | #include
13 | #include
14 |
15 | #define PI 3.1415926
16 |
17 | cv::Point prev_point;
18 |
19 | using namespace cv;
20 | using namespace std;
21 |
22 | // 해당 세터 메소드들
23 | // 누적기에 해상도 설정
24 | void LineFinder::setAccResolution(double dRho, double dTheta) {
25 | deltaRho = dRho;
26 | deltaTheta = dTheta;
27 | }
28 |
29 | // 투표 최소 개수 설정
30 | void LineFinder::setMinVote(int minv) {
31 | minVote = minv;
32 | }
33 |
34 | // 선 길이와 간격 설정
35 | void LineFinder::setLineLengthAndGap(double length, double gap) {
36 | minLength = length;
37 | maxGap = gap;
38 | }
39 |
40 | // 허프 선 세그먼트 감지를 수행하는 메소드
41 | // 확률적 허프 변환 적용
42 | std::vector LineFinder::findLines(cv::Mat& binary) {
43 | UMat gpuBinary = binary.getUMat(ACCESS_RW);
44 | lines.clear();
45 | cv::HoughLinesP(gpuBinary, lines, deltaRho, deltaTheta, minVote, minLength, maxGap);
46 | return lines;
47 | } // cv::Vec4i 벡터를 반환하고, 감지된 각 세그먼트의 시작과 마지막 점 좌표를 포함.
48 |
49 | // 위 메소드에서 감지한 선을 다음 메소드를 사용해서 그림
50 | // 영상에서 감지된 선을 그리기
51 | void LineFinder::drawDetectedLines(cv::Mat &image, cv::Scalar color) {
52 |
53 | UMat gpuImage = image.getUMat(ACCESS_RW);
54 | // 선 그리기
55 | std::vector::const_iterator it2 = lines.begin();
56 | cv::Point endPoint;
57 |
58 | while (it2 != lines.end()) {
59 | cv::Point startPoint((*it2)[0], (*it2)[1]);
60 | endPoint = cv::Point((*it2)[2], (*it2)[3]);
61 | cv::line(gpuImage, startPoint, endPoint, color, 3);
62 | ++it2;
63 | }
64 | }
65 |
--------------------------------------------------------------------------------
/linux/src/main2.cc:
--------------------------------------------------------------------------------
1 | #include
2 | #include
3 | #include
4 | #include
5 | #include
6 | #include
7 |
8 | #include
9 | #include
10 | #include
11 | #include
12 | #include
13 | #include
14 | #include
15 | #include
16 | #include
17 | #include
18 | #include
19 | #include
20 | #include
21 | #include
22 | #include
23 | #include "ets2_self_driving.h"
24 | #include "IPM.h"
25 | #include "getScreen_linux.h"
26 | #include "uinput.c"
27 |
28 | #define PI 3.1415926
29 |
30 | using namespace cv;
31 | using namespace std;
32 |
33 | void input(int, int, int);
34 | string exec(const char* cmd);
35 | int counter = 0;
36 |
37 | int main(int argc, char** argv) {
38 | if(-1 == setUinput()) {
39 | return -1;
40 | }
41 | bool detection = false;
42 | bool python_on = false;
43 | int width = 1024, height = 768;
44 | double IPM_BOTTOM_RIGHT = width+400;
45 | double IPM_BOTTOM_LEFT = -400;
46 | double IPM_RIGHT = width/2+100;
47 | double IPM_LEFT = width/2-100;
48 | int IPM_diff = 0;
49 | if(argc == 2){
50 | if(strcmp(argv[1], "--Car_Detection")){
51 | detection = true;
52 | cout<<"Car Detection Enabled"<(shmat(shmid, NULL, 0))) == (char *) -1)
90 | {
91 | printf("Error attaching shared memory id");
92 | exit(1);
93 | }
94 |
95 | int curve;
96 | key = 123464;
97 | if ((curve = shmget(key, 4, IPC_CREAT | 0666)) < 0)
98 | {
99 | printf("Error getting shared memory curve");
100 | exit(1);
101 | }
102 | // Attached shared memory
103 | if ((shared_curve = static_cast(shmat(curve, NULL, 0))) == (char *) -1)
104 | {
105 | printf("Error attaching shared memory curve");
106 | exit(1);
107 | }
108 | }
109 | }
110 | int move_mouse_pixel = 0;
111 | while (true) {
112 | auto begin = chrono::high_resolution_clock::now();
113 | // ETS2
114 | Mat image, sendImg, outputImg;
115 | int Width = 1024;
116 | int Height = 768;
117 | int Bpp = 0;
118 | std::vector Pixels;
119 |
120 | ImageFromDisplay(Pixels, Width, Height, Bpp);
121 |
122 | Mat img = Mat(Height, Width, Bpp > 24 ? CV_8UC4 : CV_8UC3, &Pixels[0]); //Mat(Size(Height, Width), Bpp > 24 ? CV_8UC4 : CV_8UC3, &Pixels[0]);
123 | cv::Rect myROI(896, 312, 1024, 768);
124 | image = img(myROI);
125 |
126 | //------------------------
127 | //cv::cvtColor(image, sendImg, CV_YUV2RGBA_NV21);
128 | cv::cvtColor(image, sendImg, CV_RGBA2RGB);
129 | //printf("%d\n", sendImg.dataend - sendImg.datastart);
130 |
131 | if(detection == true) {
132 | memcpy(shared_memory, sendImg.datastart, sendImg.dataend - sendImg.datastart);
133 | }
134 | //
135 |
136 | // Mat to GpuMat
137 | //cuda::GpuMat imageGPU;
138 | //imageGPU.upload(image);
139 |
140 | //medianBlur(image, image, 3);
141 | //cv::cuda::bilateralFilter(imageGPU, imageGPU, );
142 |
143 | IPM ipm;
144 | vector origPoints;
145 | vector dstPoints;
146 | // The 4-points at the input image
147 |
148 | origPoints.push_back(Point2f(IPM_BOTTOM_LEFT, (height-50)));
149 | origPoints.push_back(Point2f(IPM_BOTTOM_RIGHT, height-50));
150 | origPoints.push_back(Point2f(IPM_RIGHT, height/2+30));
151 | origPoints.push_back(Point2f(IPM_LEFT, height/2+30));
152 |
153 | // The 4-points correspondences in the destination image
154 |
155 | dstPoints.push_back(Point2f(0, height));
156 | dstPoints.push_back(Point2f(width, height));
157 | dstPoints.push_back(Point2f(width, 0));
158 | dstPoints.push_back(Point2f(0, 0));
159 |
160 | // IPM object
161 | ipm.setIPM(Size(width, height), Size(width, height), origPoints, dstPoints);
162 |
163 | // Process
164 | //clock_t begin = clock();
165 | ipm.applyHomography(image, outputImg);
166 | //clock_t end = clock();
167 | //double elapsed_secs = double(end - begin) / CLOCKS_PER_SEC;
168 | //printf("%.2f (ms)\r", 1000 * elapsed_secs);
169 | //ipm.drawPoints(origPoints, image);
170 |
171 | //Mat row = outputImg.row[0];
172 | cv::UMat gray;
173 | cv::UMat blur;
174 | cv::UMat sobel;
175 | cv::Mat contours;
176 | LineFinder ld; // 인스턴스 생성
177 |
178 | cv::resize(outputImg, outputImg, cv::Size(320, 240));
179 | cv::cvtColor(outputImg, gray, COLOR_RGB2GRAY);
180 | cv::blur(gray, blur, cv::Size(10, 10));
181 | cv::Sobel(blur, sobel, blur.depth(), 1, 0, 3, 0.5, 127);
182 | cv::threshold(sobel, contours, 145, 255, CV_THRESH_BINARY);
183 | //Thinning(contours, contours.rows, contours.cols);
184 | //cv::Canny(gray, contours, 125, 350);
185 |
186 |
187 | // 확률적 허프변환 파라미터 설정하기
188 |
189 | ld.setLineLengthAndGap(20, 120);
190 | ld.setMinVote(55);
191 |
192 | std::vector li = ld.findLines(contours);
193 | ld.drawDetectedLines(contours);
194 |
195 | ////////////////////////////////////////////
196 | // 자율 주행
197 |
198 | int bottom_center = 160;
199 | int sum_centerline = 0;
200 | int count_centerline = 0;
201 | int first_centerline = 0;
202 | int last_centerline = 0;
203 | double avr_center_to_left = 0;
204 | double avr_center_to_right = 0;
205 |
206 | //#pragma omp parallel for
207 | for(int i=240; i>30; i--){
208 | double center_to_right = -1;
209 | double center_to_left = -1;
210 |
211 | for (int j=0;j<150;j++) {
212 | if (contours.at(i, bottom_center+j) == 112 && center_to_right == -1) {
213 | center_to_right = j;
214 | }
215 | if (contours.at(i, bottom_center-j) == 112 && center_to_left == -1) {
216 | center_to_left = j;
217 | }
218 | }
219 | if(center_to_left!=-1 && center_to_right!=-1){
220 | int centerline = (center_to_right - center_to_left +2*bottom_center)/2;
221 | if (first_centerline == 0 ) {
222 | first_centerline = centerline;
223 | }
224 | //cv::circle(outputImg, Point(centerline, i), 1, Scalar(30, 255, 30) , 3);
225 | //cv::circle(outputImg, Point(centerline + center_to_right+20, i), 1, Scalar(255, 30, 30) , 3);
226 | //cv::circle(outputImg, Point(centerline - center_to_left+10, i), 1, Scalar(255, 30, 30) , 3);
227 | sum_centerline += centerline;
228 | avr_center_to_left = (avr_center_to_left * count_centerline + center_to_left)/count_centerline+1;
229 | avr_center_to_right = (avr_center_to_right * count_centerline + center_to_right)/count_centerline+1;
230 | last_centerline = centerline;
231 | count_centerline++;
232 | } else {
233 | }
234 | }
235 |
236 | // 컨트롤러 입력
237 | int diff = 0;
238 | if (count_centerline!=0) {
239 | diff = sum_centerline/count_centerline - bottom_center;
240 | int degree = atan2 (last_centerline - first_centerline, count_centerline) * 180 / PI;
241 | //diff = (90 - degree);
242 |
243 | move_mouse_pixel = 0 - counter + diff;
244 | cout << "Steer: "<< move_mouse_pixel << "px ";
245 | //goDirection(move_mouse_pixel);
246 | if(move_mouse_pixel == 0){
247 | if(IPM_diff > 0){
248 | IPM_RIGHT -= 1;
249 | IPM_LEFT -= 1;
250 | IPM_diff -= 1;
251 | }
252 | else if(IPM_diff < 0){
253 | IPM_RIGHT += 1;
254 | IPM_LEFT += 1;
255 | IPM_diff += 1;
256 | }
257 | else{
258 | IPM_RIGHT = width/2+100;
259 | IPM_LEFT = width/2-100;
260 | IPM_diff = 0;
261 | }
262 | }
263 | else{
264 | if (IPM_diff >= -30 && IPM_diff <= 30){
265 | IPM_RIGHT += move_mouse_pixel;
266 | IPM_LEFT += move_mouse_pixel;
267 | if(move_mouse_pixel > 0){
268 | IPM_diff++;
269 | }
270 | else{
271 | IPM_diff--;
272 | }
273 | }
274 | }
275 | moveMouse(move_mouse_pixel);
276 | int road_curve = (int)((sum_centerline/count_centerline-bottom_center));
277 |
278 | if(detection == true) {
279 | memcpy(shared_curve, &road_curve, sizeof(int));
280 | }
281 |
282 | counter = diff;/*
283 |
284 |
285 | if (abs(move_mouse_pixel)< 5) {
286 | goDirection(0 - counter/3);
287 | counter -= counter/3;
288 | } else if (abs(move_mouse_pixel) < 4) {
289 | goDirection(0 - counter/2);
290 | counter -= counter/2;
291 | } else if (abs (move_mouse_pixel) < 2) {
292 | goDirection(0 - counter);
293 | counter = 0;
294 | } else {
295 | goDirection(move_mouse_pixel);
296 | counter += move_mouse_pixel;
297 | }
298 | */
299 |
300 | } else {}
301 |
302 |
303 | ////////////////////////////////////////////
304 |
305 | //cv::cvtColor(contours, contours, COLOR_GRAY2RGB);
306 | imshow("Road", outputImg);
307 | imshow("Lines", contours);
308 | moveWindow("Lines", 200, 400);
309 | moveWindow("Road", 530, 400);
310 | waitKey(1);
311 |
312 | auto end = chrono::high_resolution_clock::now();
313 | auto dur = end - begin;
314 | auto ms = std::chrono::duration_cast(dur).count();
315 | ms++;
316 | //cout << 1000 / ms << "fps avr:" << 1000 / (sum / (++i)) << endl;
317 | cout << 1000 / ms << "fps" << endl;
318 |
319 | }
320 | return 0;
321 | }
322 |
323 |
324 |
325 |
326 | std::string exec(const char* cmd) {
327 | std::array buffer;
328 | std::string result;
329 | std::shared_ptr pipe(popen(cmd, "r"), pclose);
330 | if (!pipe)
331 | throw std::runtime_error("popen() failed!");
332 | while (!feof(pipe.get())) {
333 | if (fgets(buffer.data(), 128, pipe.get()) != NULL)
334 | result += buffer.data();
335 | }
336 | return result;
337 | }
338 |
--------------------------------------------------------------------------------
/linux/src/uinput.c:
--------------------------------------------------------------------------------
1 | #include
2 | #include "uinput.h"
3 |
4 | void setEventAndWrite(__u16 type, __u16 code, __s32 value)
5 | {
6 | ev.type=type;
7 | ev.code=code;
8 | ev.value=value;
9 | if(write(fd, &ev, sizeof(struct input_event)) < 0)
10 | die("error: write");
11 | }
12 |
13 | int setUinput() {
14 | fd = open("/dev/uinput", O_WRONLY | O_NONBLOCK);
15 | if(fd < 0) {
16 | fd=open("/dev/input/uinput",O_WRONLY|O_NONBLOCK);
17 | if(fd<0){
18 | die("error: Can't open an uinput file. see #17.(https://github.com/bethesirius/ChosunTruck/issues/17)");
19 | return -1;
20 | }
21 | }
22 | if(ioctl(fd, UI_SET_EVBIT, EV_KEY) < 0) {
23 | die("error: ioctl");
24 | return -1;
25 | }
26 | if(ioctl(fd, UI_SET_KEYBIT, BTN_LEFT) < 0) {
27 | die("error: ioctl");
28 | return -1;
29 | }
30 | if(ioctl(fd, UI_SET_KEYBIT, KEY_TAB) < 0) {
31 | die("error: ioctl");
32 | return -1;
33 | }
34 | if(ioctl(fd, UI_SET_KEYBIT, KEY_ENTER) < 0) {
35 | die("error: ioctl");
36 | return -1;
37 | }
38 | if(ioctl(fd, UI_SET_KEYBIT, KEY_LEFTSHIFT) < 0) {
39 | die("error: ioctl");
40 | return -1;
41 | }
42 | if(ioctl(fd, UI_SET_EVBIT, EV_REL) < 0) {
43 | die("error: ioctl");
44 | return -1;
45 | }
46 | if(ioctl(fd, UI_SET_RELBIT, REL_X) < 0) {
47 | die("error: ioctl");
48 | return -1;
49 | }
50 | if(ioctl(fd, UI_SET_RELBIT, REL_Y) < 0) {
51 | die("error: ioctl");
52 | return -1;
53 | }
54 |
55 | memset(&uidev, 0, sizeof(uidev));
56 | snprintf(uidev.name, UINPUT_MAX_NAME_SIZE, "uinput-virtualMouse");
57 | uidev.id.bustype = BUS_USB;
58 | uidev.id.vendor = 0x1;
59 | uidev.id.product = 0x1;
60 | uidev.id.version = 1;
61 |
62 | if(write(fd, &uidev, sizeof(uidev)) < 0) {
63 | die("error: write");
64 | return -1;
65 | }
66 | if(ioctl(fd, UI_DEV_CREATE) < 0) {
67 | die("error: ioctl");
68 | return -1;
69 | }
70 |
71 | memset(&ev, 0, sizeof(struct input_event));
72 | setEventAndWrite(EV_REL,REL_X,1);
73 | setEventAndWrite(EV_REL,REL_Y,1);
74 | setEventAndWrite(EV_SYN,0,0);
75 | return 0;
76 | }
77 |
78 | void moveMouse(int x) {
79 | int dx=x;
80 | setEventAndWrite(EV_REL,REL_X,dx);
81 | setEventAndWrite(EV_SYN,0,0);
82 | }
83 |
84 |
85 |
--------------------------------------------------------------------------------
/linux/src/uinput.h:
--------------------------------------------------------------------------------
1 | #include
2 | #include
3 | #include
4 | #include
5 | #include
6 | #include
7 | #include
8 | #include
9 | #include
10 | #include
11 | #include
12 |
13 | #define die(str, args...) do { \
14 | perror(str); \
15 | exit(EXIT_FAILURE); \
16 | } while(0)
17 |
18 | static int fd;//uinput fd
19 | static struct uinput_user_dev uidev;
20 | static struct input_event ev;
21 |
22 | int setUinput();
23 | void moveMouse(int);
24 |
25 |
--------------------------------------------------------------------------------
/linux/tensorbox/Car_detection.py:
--------------------------------------------------------------------------------
1 | import tensorflow as tf
2 | import os
3 | import json
4 | import subprocess
5 | import sysv_ipc
6 | import struct
7 |
8 | from scipy.misc import imread, imresize
9 | from scipy import misc
10 |
11 | from train import build_forward
12 | from utils.annolist import AnnotationLib as al
13 | from utils.train_utils import add_rectangles, rescale_boxes
14 | from pymouse import PyMouse
15 |
16 | import cv2
17 | import argparse
18 | import time
19 | import numpy as np
20 |
21 | def get_image_dir(args):
22 | weights_iteration = int(args.weights.split('-')[-1])
23 | expname = '_' + args.expname if args.expname else ''
24 | image_dir = '%s/images_%s_%d%s' % (os.path.dirname(args.weights), os.path.basename(args.test_boxes)[:-5], weights_iteration, expname)
25 | return image_dir
26 |
27 | def get_results(args, H):
28 | tf.reset_default_graph()
29 | x_in = tf.placeholder(tf.float32, name='x_in', shape=[H['image_height'], H['image_width'], 3])
30 | if H['use_rezoom']:
31 | pred_boxes, pred_logits, pred_confidences, pred_confs_deltas, pred_boxes_deltas = build_forward(H, tf.expand_dims(x_in, 0), 'test', reuse=None)
32 | grid_area = H['grid_height'] * H['grid_width']
33 | pred_confidences = tf.reshape(tf.nn.softmax(tf.reshape(pred_confs_deltas, [grid_area * H['rnn_len'], 2])), [grid_area, H['rnn_len'], 2])
34 | if H['reregress']:
35 | pred_boxes = pred_boxes + pred_boxes_deltas
36 | else:
37 | pred_boxes, pred_logits, pred_confidences = build_forward(H, tf.expand_dims(x_in, 0), 'test', reuse=None)
38 | saver = tf.train.Saver()
39 | with tf.Session() as sess:
40 | sess.run(tf.initialize_all_variables())
41 | saver.restore(sess, args.weights)
42 |
43 | pred_annolist = al.AnnoList()
44 |
45 | data_dir = os.path.dirname(args.test_boxes)
46 | image_dir = get_image_dir(args)
47 | subprocess.call('mkdir -p %s' % image_dir, shell=True)
48 |
49 | memory = sysv_ipc.SharedMemory(123463)
50 | memory2 = sysv_ipc.SharedMemory(123464)
51 | size = 768, 1024, 3
52 |
53 | pedal = PyMouse()
54 | pedal.press(1)
55 | road_center = 320
56 | while True:
57 | cv2.waitKey(1)
58 | frameCount = bytearray(memory.read())
59 | curve = bytearray(memory2.read())
60 | curve = str(struct.unpack('i',curve)[0])
61 | m = np.array(frameCount, dtype=np.uint8)
62 | orig_img = m.reshape(size)
63 |
64 | img = imresize(orig_img, (H["image_height"], H["image_width"]), interp='cubic')
65 | feed = {x_in: img}
66 | (np_pred_boxes, np_pred_confidences) = sess.run([pred_boxes, pred_confidences], feed_dict=feed)
67 | pred_anno = al.Annotation()
68 |
69 | new_img, rects = add_rectangles(H, [img], np_pred_confidences, np_pred_boxes,
70 | use_stitching=True, rnn_len=H['rnn_len'], min_conf=args.min_conf, tau=args.tau, show_suppressed=args.show_suppressed)
71 | flag = 0
72 | road_center = 320 + int(curve)
73 | print(road_center)
74 | for rect in rects:
75 | print(rect.x1, rect.x2, rect.y2)
76 | if (rect.x1 < road_center and rect.x2 > road_center and rect.y2 > 200) and (rect.x2 - rect.x1 > 30):
77 | flag = 1
78 |
79 | if flag is 1:
80 | pedal.press(2)
81 | print("break!")
82 | else:
83 | pedal.release(2)
84 | pedal.press(1)
85 | print("acceleration!")
86 |
87 | pred_anno.rects = rects
88 | pred_anno.imagePath = os.path.abspath(data_dir)
89 | pred_anno = rescale_boxes((H["image_height"], H["image_width"]), pred_anno, orig_img.shape[0], orig_img.shape[1])
90 | pred_annolist.append(pred_anno)
91 |
92 | cv2.imshow('.jpg', new_img)
93 |
94 | return none;
95 |
96 | def main():
97 | parser = argparse.ArgumentParser()
98 | parser.add_argument('--weights', default='tensorbox/output/overfeat_rezoom_2017_02_09_13.28/save.ckpt-100000')
99 | parser.add_argument('--expname', default='')
100 | parser.add_argument('--test_boxes', default='default')
101 | parser.add_argument('--gpu', default=0)
102 | parser.add_argument('--logdir', default='output')
103 | parser.add_argument('--iou_threshold', default=0.5, type=float)
104 | parser.add_argument('--tau', default=0.25, type=float)
105 | parser.add_argument('--min_conf', default=0.2, type=float)
106 | parser.add_argument('--show_suppressed', default=True, type=bool)
107 | args = parser.parse_args()
108 | os.environ['CUDA_VISIBLE_DEVICES'] = str(args.gpu)
109 | hypes_file = '%s/hypes.json' % os.path.dirname(args.weights)
110 | with open(hypes_file, 'r') as f:
111 | H = json.load(f)
112 | expname = args.expname + '_' if args.expname else ''
113 | pred_boxes = '%s.%s%s' % (args.weights, expname, os.path.basename(args.test_boxes))
114 |
115 | get_results(args, H)
116 |
117 | if __name__ == '__main__':
118 | main()
119 |
--------------------------------------------------------------------------------
/linux/tensorbox/data/inception_v1.ckpt:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/bethesirius/ChosunTruck/889644385ce57f971ec2921f006fbb0a167e6f1e/linux/tensorbox/data/inception_v1.ckpt
--------------------------------------------------------------------------------
/linux/tensorbox/license/TENSORBOX_LICENSE.txt:
--------------------------------------------------------------------------------
1 | Copyright (c) 2016 "The Contributors"
2 |
3 |
4 | Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
5 |
6 | The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
7 |
8 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
9 |
--------------------------------------------------------------------------------
/linux/tensorbox/license/TENSORFLOW_LICENSE.txt:
--------------------------------------------------------------------------------
1 | Copyright 2015 The TensorFlow Authors. All rights reserved.
2 |
3 | Apache License
4 | Version 2.0, January 2004
5 | http://www.apache.org/licenses/
6 |
7 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
8 |
9 | 1. Definitions.
10 |
11 | "License" shall mean the terms and conditions for use, reproduction,
12 | and distribution as defined by Sections 1 through 9 of this document.
13 |
14 | "Licensor" shall mean the copyright owner or entity authorized by
15 | the copyright owner that is granting the License.
16 |
17 | "Legal Entity" shall mean the union of the acting entity and all
18 | other entities that control, are controlled by, or are under common
19 | control with that entity. For the purposes of this definition,
20 | "control" means (i) the power, direct or indirect, to cause the
21 | direction or management of such entity, whether by contract or
22 | otherwise, or (ii) ownership of fifty percent (50%) or more of the
23 | outstanding shares, or (iii) beneficial ownership of such entity.
24 |
25 | "You" (or "Your") shall mean an individual or Legal Entity
26 | exercising permissions granted by this License.
27 |
28 | "Source" form shall mean the preferred form for making modifications,
29 | including but not limited to software source code, documentation
30 | source, and configuration files.
31 |
32 | "Object" form shall mean any form resulting from mechanical
33 | transformation or translation of a Source form, including but
34 | not limited to compiled object code, generated documentation,
35 | and conversions to other media types.
36 |
37 | "Work" shall mean the work of authorship, whether in Source or
38 | Object form, made available under the License, as indicated by a
39 | copyright notice that is included in or attached to the work
40 | (an example is provided in the Appendix below).
41 |
42 | "Derivative Works" shall mean any work, whether in Source or Object
43 | form, that is based on (or derived from) the Work and for which the
44 | editorial revisions, annotations, elaborations, or other modifications
45 | represent, as a whole, an original work of authorship. For the purposes
46 | of this License, Derivative Works shall not include works that remain
47 | separable from, or merely link (or bind by name) to the interfaces of,
48 | the Work and Derivative Works thereof.
49 |
50 | "Contribution" shall mean any work of authorship, including
51 | the original version of the Work and any modifications or additions
52 | to that Work or Derivative Works thereof, that is intentionally
53 | submitted to Licensor for inclusion in the Work by the copyright owner
54 | or by an individual or Legal Entity authorized to submit on behalf of
55 | the copyright owner. For the purposes of this definition, "submitted"
56 | means any form of electronic, verbal, or written communication sent
57 | to the Licensor or its representatives, including but not limited to
58 | communication on electronic mailing lists, source code control systems,
59 | and issue tracking systems that are managed by, or on behalf of, the
60 | Licensor for the purpose of discussing and improving the Work, but
61 | excluding communication that is conspicuously marked or otherwise
62 | designated in writing by the copyright owner as "Not a Contribution."
63 |
64 | "Contributor" shall mean Licensor and any individual or Legal Entity
65 | on behalf of whom a Contribution has been received by Licensor and
66 | subsequently incorporated within the Work.
67 |
68 | 2. Grant of Copyright License. Subject to the terms and conditions of
69 | this License, each Contributor hereby grants to You a perpetual,
70 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable
71 | copyright license to reproduce, prepare Derivative Works of,
72 | publicly display, publicly perform, sublicense, and distribute the
73 | Work and such Derivative Works in Source or Object form.
74 |
75 | 3. Grant of Patent License. Subject to the terms and conditions of
76 | this License, each Contributor hereby grants to You a perpetual,
77 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable
78 | (except as stated in this section) patent license to make, have made,
79 | use, offer to sell, sell, import, and otherwise transfer the Work,
80 | where such license applies only to those patent claims licensable
81 | by such Contributor that are necessarily infringed by their
82 | Contribution(s) alone or by combination of their Contribution(s)
83 | with the Work to which such Contribution(s) was submitted. If You
84 | institute patent litigation against any entity (including a
85 | cross-claim or counterclaim in a lawsuit) alleging that the Work
86 | or a Contribution incorporated within the Work constitutes direct
87 | or contributory patent infringement, then any patent licenses
88 | granted to You under this License for that Work shall terminate
89 | as of the date such litigation is filed.
90 |
91 | 4. Redistribution. You may reproduce and distribute copies of the
92 | Work or Derivative Works thereof in any medium, with or without
93 | modifications, and in Source or Object form, provided that You
94 | meet the following conditions:
95 |
96 | (a) You must give any other recipients of the Work or
97 | Derivative Works a copy of this License; and
98 |
99 | (b) You must cause any modified files to carry prominent notices
100 | stating that You changed the files; and
101 |
102 | (c) You must retain, in the Source form of any Derivative Works
103 | that You distribute, all copyright, patent, trademark, and
104 | attribution notices from the Source form of the Work,
105 | excluding those notices that do not pertain to any part of
106 | the Derivative Works; and
107 |
108 | (d) If the Work includes a "NOTICE" text file as part of its
109 | distribution, then any Derivative Works that You distribute must
110 | include a readable copy of the attribution notices contained
111 | within such NOTICE file, excluding those notices that do not
112 | pertain to any part of the Derivative Works, in at least one
113 | of the following places: within a NOTICE text file distributed
114 | as part of the Derivative Works; within the Source form or
115 | documentation, if provided along with the Derivative Works; or,
116 | within a display generated by the Derivative Works, if and
117 | wherever such third-party notices normally appear. The contents
118 | of the NOTICE file are for informational purposes only and
119 | do not modify the License. You may add Your own attribution
120 | notices within Derivative Works that You distribute, alongside
121 | or as an addendum to the NOTICE text from the Work, provided
122 | that such additional attribution notices cannot be construed
123 | as modifying the License.
124 |
125 | You may add Your own copyright statement to Your modifications and
126 | may provide additional or different license terms and conditions
127 | for use, reproduction, or distribution of Your modifications, or
128 | for any such Derivative Works as a whole, provided Your use,
129 | reproduction, and distribution of the Work otherwise complies with
130 | the conditions stated in this License.
131 |
132 | 5. Submission of Contributions. Unless You explicitly state otherwise,
133 | any Contribution intentionally submitted for inclusion in the Work
134 | by You to the Licensor shall be under the terms and conditions of
135 | this License, without any additional terms or conditions.
136 | Notwithstanding the above, nothing herein shall supersede or modify
137 | the terms of any separate license agreement you may have executed
138 | with Licensor regarding such Contributions.
139 |
140 | 6. Trademarks. This License does not grant permission to use the trade
141 | names, trademarks, service marks, or product names of the Licensor,
142 | except as required for reasonable and customary use in describing the
143 | origin of the Work and reproducing the content of the NOTICE file.
144 |
145 | 7. Disclaimer of Warranty. Unless required by applicable law or
146 | agreed to in writing, Licensor provides the Work (and each
147 | Contributor provides its Contributions) on an "AS IS" BASIS,
148 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
149 | implied, including, without limitation, any warranties or conditions
150 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
151 | PARTICULAR PURPOSE. You are solely responsible for determining the
152 | appropriateness of using or redistributing the Work and assume any
153 | risks associated with Your exercise of permissions under this License.
154 |
155 | 8. Limitation of Liability. In no event and under no legal theory,
156 | whether in tort (including negligence), contract, or otherwise,
157 | unless required by applicable law (such as deliberate and grossly
158 | negligent acts) or agreed to in writing, shall any Contributor be
159 | liable to You for damages, including any direct, indirect, special,
160 | incidental, or consequential damages of any character arising as a
161 | result of this License or out of the use or inability to use the
162 | Work (including but not limited to damages for loss of goodwill,
163 | work stoppage, computer failure or malfunction, or any and all
164 | other commercial damages or losses), even if such Contributor
165 | has been advised of the possibility of such damages.
166 |
167 | 9. Accepting Warranty or Additional Liability. While redistributing
168 | the Work or Derivative Works thereof, You may choose to offer,
169 | and charge a fee for, acceptance of support, warranty, indemnity,
170 | or other liability obligations and/or rights consistent with this
171 | License. However, in accepting such obligations, You may act only
172 | on Your own behalf and on Your sole responsibility, not on behalf
173 | of any other Contributor, and only if You agree to indemnify,
174 | defend, and hold each Contributor harmless for any liability
175 | incurred by, or claims asserted against, such Contributor by reason
176 | of your accepting any such warranty or additional liability.
177 |
178 | END OF TERMS AND CONDITIONS
179 |
180 | APPENDIX: How to apply the Apache License to your work.
181 |
182 | To apply the Apache License to your work, attach the following
183 | boilerplate notice, with the fields enclosed by brackets "[]"
184 | replaced with your own identifying information. (Don't include
185 | the brackets!) The text should be enclosed in the appropriate
186 | comment syntax for the file format. We also recommend that a
187 | file or class name and description of purpose be included on the
188 | same "printed page" as the copyright notice for easier
189 | identification within third-party archives.
190 |
191 | Copyright 2015, The TensorFlow Authors.
192 |
193 | Licensed under the Apache License, Version 2.0 (the "License");
194 | you may not use this file except in compliance with the License.
195 | You may obtain a copy of the License at
196 |
197 | http://www.apache.org/licenses/LICENSE-2.0
198 |
199 | Unless required by applicable law or agreed to in writing, software
200 | distributed under the License is distributed on an "AS IS" BASIS,
201 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
202 | See the License for the specific language governing permissions and
203 | limitations under the License.
204 |
--------------------------------------------------------------------------------
/linux/tensorbox/output/overfeat_rezoom_2017_02_09_13.28/checkpoint:
--------------------------------------------------------------------------------
1 | model_checkpoint_path: "save.ckpt-100000"
2 | all_model_checkpoint_paths: "save.ckpt-100000"
3 |
--------------------------------------------------------------------------------
/linux/tensorbox/output/overfeat_rezoom_2017_02_09_13.28/hypes.json:
--------------------------------------------------------------------------------
1 | {
2 | "deconv": false,
3 | "lstm_size": 500,
4 | "later_feat_channels": 832,
5 | "slim_basename": "InceptionV1",
6 | "image_width": 640,
7 | "rezoom_h_coords": [
8 | -0.25,
9 | 0.25
10 | ],
11 | "grid_width": 20,
12 | "exp_name": "overfeat_rezoom",
13 | "use_lstm": false,
14 | "region_size": 32,
15 | "num_lstm_layers": 2,
16 | "focus_size": 1.8,
17 | "avg_pool_size": 5,
18 | "early_feat_channels": 256,
19 | "grid_height": 15,
20 | "use_rezoom": true,
21 | "slim_top_lname": "Mixed_5b",
22 | "rezoom_w_coords": [
23 | -0.25,
24 | 0.25
25 | ],
26 | "rezoom_change_loss": "center",
27 | "batch_size": 1,
28 | "reregress": true,
29 | "data": {
30 | "test_idl": "./data/truckdata/val.json",
31 | "truncate_data": false,
32 | "train_idl": "./data/truckdata/train.json"
33 | },
34 | "num_classes": 2,
35 | "logging": {
36 | "save_iter": 10000,
37 | "display_iter": 50
38 | },
39 | "rnn_len": 1,
40 | "solver": {
41 | "opt": "RMS",
42 | "rnd_seed": 1,
43 | "epsilon": 1e-05,
44 | "learning_rate": 0.001,
45 | "use_jitter": false,
46 | "hungarian_iou": 0.25,
47 | "weights": "",
48 | "learning_rate_step": 33000,
49 | "gpu": 0,
50 | "head_weights": [
51 | 1.0,
52 | 0.1
53 | ]
54 | },
55 | "clip_norm": 1.0,
56 | "image_height": 480,
57 | "save_dir": "output/overfeat_rezoom_2017_02_09_13.28",
58 | "slim_attention_lname": "Mixed_3b",
59 | "slim_ckpt": "inception_v1.ckpt",
60 | "biggest_box_px": 10000
61 | }
--------------------------------------------------------------------------------
/linux/tensorbox/output/overfeat_rezoom_2017_02_09_13.28/save.ckpt-100000.data-00000-of-00001:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/bethesirius/ChosunTruck/889644385ce57f971ec2921f006fbb0a167e6f1e/linux/tensorbox/output/overfeat_rezoom_2017_02_09_13.28/save.ckpt-100000.data-00000-of-00001
--------------------------------------------------------------------------------
/linux/tensorbox/output/overfeat_rezoom_2017_02_09_13.28/save.ckpt-100000.index:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/bethesirius/ChosunTruck/889644385ce57f971ec2921f006fbb0a167e6f1e/linux/tensorbox/output/overfeat_rezoom_2017_02_09_13.28/save.ckpt-100000.index
--------------------------------------------------------------------------------
/linux/tensorbox/pymouse/__init__.py:
--------------------------------------------------------------------------------
1 | #Copyright 2013 Paul Barton
2 | #
3 | #This program is free software: you can redistribute it and/or modify
4 | #it under the terms of the GNU General Public License as published by
5 | #the Free Software Foundation, either version 3 of the License, or
6 | #(at your option) any later version.
7 | #
8 | #This program is distributed in the hope that it will be useful,
9 | #but WITHOUT ANY WARRANTY; without even the implied warranty of
10 | #MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 | #GNU General Public License for more details.
12 | #
13 | #You should have received a copy of the GNU General Public License
14 | #along with this program. If not, see .
15 |
16 | """
17 | The goal of PyMouse is to have a cross-platform way to control the mouse.
18 | PyMouse should work on Windows, Mac and any Unix that has xlib.
19 |
20 | PyMouse is a part of PyUserInput, along with PyKeyboard, for more information
21 | about this project, see:
22 | http://github.com/SavinaRoja/PyUserInput
23 |
24 | PyMouse was originally developed by Pepijn de Vos. For the original repository,
25 | see:
26 | https://github.com/pepijndevos/PyMouse
27 | """
28 |
29 | import sys
30 |
31 | if sys.platform.startswith('java'):
32 | from .java_ import PyMouse
33 |
34 | elif sys.platform == 'darwin':
35 | from .mac import PyMouse, PyMouseEvent
36 |
37 | elif sys.platform == 'win32':
38 | from .windows import PyMouse, PyMouseEvent
39 |
40 | else:
41 | from .x11 import PyMouse, PyMouseEvent
42 |
43 |
--------------------------------------------------------------------------------
/linux/tensorbox/pymouse/base.py:
--------------------------------------------------------------------------------
1 | #Copyright 2013 Paul Barton
2 | #
3 | #This program is free software: you can redistribute it and/or modify
4 | #it under the terms of the GNU General Public License as published by
5 | #the Free Software Foundation, either version 3 of the License, or
6 | #(at your option) any later version.
7 | #
8 | #This program is distributed in the hope that it will be useful,
9 | #but WITHOUT ANY WARRANTY; without even the implied warranty of
10 | #MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 | #GNU General Public License for more details.
12 | #
13 | #You should have received a copy of the GNU General Public License
14 | #along with this program. If not, see .
15 |
16 | """
17 | The goal of PyMouse is to have a cross-platform way to control the mouse.
18 | PyMouse should work on Windows, Mac and any Unix that has xlib.
19 |
20 | As the base file, this provides a rough operational model along with the
21 | framework to be extended by each platform.
22 | """
23 |
24 | from threading import Thread
25 |
26 |
27 | class ScrollSupportError(Exception):
28 | pass
29 |
30 |
31 | class PyMouseMeta(object):
32 |
33 | def press(self, x, y, button=1):
34 | """
35 | Press the mouse on a given x, y and button.
36 | Button is defined as 1 = left, 2 = right, 3 = middle.
37 | """
38 |
39 | raise NotImplementedError
40 |
41 | def release(self, x, y, button=1):
42 | """
43 | Release the mouse on a given x, y and button.
44 | Button is defined as 1 = left, 2 = right, 3 = middle.
45 | """
46 |
47 | raise NotImplementedError
48 |
49 | def click(self, x, y, button=1, n=1):
50 | """
51 | Click a mouse button n times on a given x, y.
52 | Button is defined as 1 = left, 2 = right, 3 = middle.
53 | """
54 |
55 | for i in range(n):
56 | self.press(x, y, button)
57 | self.release(x, y, button)
58 |
59 | def scroll(self, vertical=None, horizontal=None, depth=None):
60 | """
61 | Generates mouse scrolling events in up to three dimensions: vertical,
62 | horizontal, and depth (Mac-only). Values for these arguments may be
63 | positive or negative numbers (float or int). Refer to the following:
64 | Vertical: + Up, - Down
65 | Horizontal: + Right, - Left
66 | Depth: + Rise (out of display), - Dive (towards display)
67 |
68 | Dynamic scrolling, which is used Windows and Mac platforms, is not
69 | implemented at this time due to an inability to test Mac code. The
70 | events generated by this code will thus be discrete units of scrolling
71 | "lines". The user is advised to take care at all times with scrolling
72 | automation as scrolling event consumption is relatively un-standardized.
73 |
74 | Float values will be coerced to integers.
75 | """
76 |
77 | raise NotImplementedError
78 |
79 | def move(self, x, y):
80 | """Move the mouse to a given x and y"""
81 |
82 | raise NotImplementedError
83 |
84 | def drag(self, x, y):
85 | """Drag the mouse to a given x and y.
86 | A Drag is a Move where the mouse key is held down."""
87 |
88 | raise NotImplementedError
89 |
90 | def position(self):
91 | """
92 | Get the current mouse position in pixels.
93 | Returns a tuple of 2 integers
94 | """
95 |
96 | raise NotImplementedError
97 |
98 | def screen_size(self):
99 | """
100 | Get the current screen size in pixels.
101 | Returns a tuple of 2 integers
102 | """
103 |
104 | raise NotImplementedError
105 |
106 |
107 | class PyMouseEventMeta(Thread):
108 | def __init__(self, capture=False, capture_move=False):
109 | Thread.__init__(self)
110 | self.daemon = True
111 | self.capture = capture
112 | self.capture_move = capture_move
113 | self.state = True
114 |
115 | def stop(self):
116 | self.state = False
117 |
118 | def click(self, x, y, button, press):
119 | """Subclass this method with your click event handler"""
120 | pass
121 |
122 | def move(self, x, y):
123 | """Subclass this method with your move event handler"""
124 | pass
125 |
126 | def scroll(self, x, y, vertical, horizontal):
127 | """
128 | Subclass this method with your scroll event handler
129 | Vertical: + Up, - Down
130 | Horizontal: + Right, - Left
131 | """
132 | pass
133 |
--------------------------------------------------------------------------------
/linux/tensorbox/pymouse/java_.py:
--------------------------------------------------------------------------------
1 | #Copyright 2013 Paul Barton
2 | #
3 | #This program is free software: you can redistribute it and/or modify
4 | #it under the terms of the GNU General Public License as published by
5 | #the Free Software Foundation, either version 3 of the License, or
6 | #(at your option) any later version.
7 | #
8 | #This program is distributed in the hope that it will be useful,
9 | #but WITHOUT ANY WARRANTY; without even the implied warranty of
10 | #MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 | #GNU General Public License for more details.
12 | #
13 | #You should have received a copy of the GNU General Public License
14 | #along with this program. If not, see .
15 |
16 | from java.awt import Robot, Toolkit
17 | from java.awt.event import InputEvent
18 | from java.awt.MouseInfo import getPointerInfo
19 | from .base import PyMouseMeta
20 |
21 | r = Robot()
22 |
23 | class PyMouse(PyMouseMeta):
24 | def press(self, x, y, button = 1):
25 | button_list = [None, InputEvent.BUTTON1_MASK, InputEvent.BUTTON3_MASK, InputEvent.BUTTON2_MASK]
26 | self.move(x, y)
27 | r.mousePress(button_list[button])
28 |
29 | def release(self, x, y, button = 1):
30 | button_list = [None, InputEvent.BUTTON1_MASK, InputEvent.BUTTON3_MASK, InputEvent.BUTTON2_MASK]
31 | self.move(x, y)
32 | r.mouseRelease(button_list[button])
33 |
34 | def move(self, x, y):
35 | r.mouseMove(x, y)
36 |
37 | def position(self):
38 | loc = getPointerInfo().getLocation()
39 | return loc.getX, loc.getY
40 |
41 | def screen_size(self):
42 | dim = Toolkit.getDefaultToolkit().getScreenSize()
43 | return dim.getWidth(), dim.getHeight()
44 |
--------------------------------------------------------------------------------
/linux/tensorbox/pymouse/mac.py:
--------------------------------------------------------------------------------
1 | #Copyright 2013 Paul Barton
2 | #
3 | #This program is free software: you can redistribute it and/or modify
4 | #it under the terms of the GNU General Public License as published by
5 | #the Free Software Foundation, either version 3 of the License, or
6 | #(at your option) any later version.
7 | #
8 | #This program is distributed in the hope that it will be useful,
9 | #but WITHOUT ANY WARRANTY; without even the implied warranty of
10 | #MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 | #GNU General Public License for more details.
12 | #
13 | #You should have received a copy of the GNU General Public License
14 | #along with this program. If not, see .
15 |
16 | import Quartz
17 | from AppKit import NSEvent, NSScreen
18 | from .base import PyMouseMeta, PyMouseEventMeta
19 |
20 | pressID = [None, Quartz.kCGEventLeftMouseDown,
21 | Quartz.kCGEventRightMouseDown, Quartz.kCGEventOtherMouseDown]
22 | releaseID = [None, Quartz.kCGEventLeftMouseUp,
23 | Quartz.kCGEventRightMouseUp, Quartz.kCGEventOtherMouseUp]
24 |
25 |
26 | class PyMouse(PyMouseMeta):
27 |
28 | def press(self, x, y, button=1):
29 | event = Quartz.CGEventCreateMouseEvent(None,
30 | pressID[button],
31 | (x, y),
32 | button - 1)
33 | Quartz.CGEventPost(Quartz.kCGHIDEventTap, event)
34 |
35 | def release(self, x, y, button=1):
36 | event = Quartz.CGEventCreateMouseEvent(None,
37 | releaseID[button],
38 | (x, y),
39 | button - 1)
40 | Quartz.CGEventPost(Quartz.kCGHIDEventTap, event)
41 |
42 | def move(self, x, y):
43 | move = Quartz.CGEventCreateMouseEvent(None, Quartz.kCGEventMouseMoved, (x, y), 0)
44 | Quartz.CGEventPost(Quartz.kCGHIDEventTap, move)
45 |
46 | def drag(self, x, y):
47 | drag = Quartz.CGEventCreateMouseEvent(None, Quartz.kCGEventLeftMouseDragged, (x, y), 0)
48 | Quartz.CGEventPost(Quartz.kCGHIDEventTap, drag)
49 |
50 | def position(self):
51 | loc = NSEvent.mouseLocation()
52 | return loc.x, Quartz.CGDisplayPixelsHigh(0) - loc.y
53 |
54 | def screen_size(self):
55 | return NSScreen.mainScreen().frame().size.width, NSScreen.mainScreen().frame().size.height
56 |
57 | def scroll(self, vertical=None, horizontal=None, depth=None):
58 | #Local submethod for generating Mac scroll events in one axis at a time
59 | def scroll_event(y_move=0, x_move=0, z_move=0, n=1):
60 | for _ in range(abs(n)):
61 | scrollWheelEvent = Quartz.CGEventCreateScrollWheelEvent(
62 | None, # No source
63 | Quartz.kCGScrollEventUnitLine, # Unit of measurement is lines
64 | 3, # Number of wheels(dimensions)
65 | y_move,
66 | x_move,
67 | z_move)
68 | Quartz.CGEventPost(Quartz.kCGHIDEventTap, scrollWheelEvent)
69 |
70 | #Execute vertical then horizontal then depth scrolling events
71 | if vertical is not None:
72 | vertical = int(vertical)
73 | if vertical == 0: # Do nothing with 0 distance
74 | pass
75 | elif vertical > 0: # Scroll up if positive
76 | scroll_event(y_move=1, n=vertical)
77 | else: # Scroll down if negative
78 | scroll_event(y_move=-1, n=abs(vertical))
79 | if horizontal is not None:
80 | horizontal = int(horizontal)
81 | if horizontal == 0: # Do nothing with 0 distance
82 | pass
83 | elif horizontal > 0: # Scroll right if positive
84 | scroll_event(x_move=1, n=horizontal)
85 | else: # Scroll left if negative
86 | scroll_event(x_move=-1, n=abs(horizontal))
87 | if depth is not None:
88 | depth = int(depth)
89 | if depth == 0: # Do nothing with 0 distance
90 | pass
91 | elif vertical > 0: # Scroll "out" if positive
92 | scroll_event(z_move=1, n=depth)
93 | else: # Scroll "in" if negative
94 | scroll_event(z_move=-1, n=abs(depth))
95 |
96 |
97 | class PyMouseEvent(PyMouseEventMeta):
98 | def run(self):
99 | tap = Quartz.CGEventTapCreate(
100 | Quartz.kCGSessionEventTap,
101 | Quartz.kCGHeadInsertEventTap,
102 | Quartz.kCGEventTapOptionDefault,
103 | Quartz.CGEventMaskBit(Quartz.kCGEventMouseMoved) |
104 | Quartz.CGEventMaskBit(Quartz.kCGEventLeftMouseDown) |
105 | Quartz.CGEventMaskBit(Quartz.kCGEventLeftMouseUp) |
106 | Quartz.CGEventMaskBit(Quartz.kCGEventRightMouseDown) |
107 | Quartz.CGEventMaskBit(Quartz.kCGEventRightMouseUp) |
108 | Quartz.CGEventMaskBit(Quartz.kCGEventOtherMouseDown) |
109 | Quartz.CGEventMaskBit(Quartz.kCGEventOtherMouseUp),
110 | self.handler,
111 | None)
112 |
113 | loopsource = Quartz.CFMachPortCreateRunLoopSource(None, tap, 0)
114 | loop = Quartz.CFRunLoopGetCurrent()
115 | Quartz.CFRunLoopAddSource(loop, loopsource, Quartz.kCFRunLoopDefaultMode)
116 | Quartz.CGEventTapEnable(tap, True)
117 |
118 | while self.state:
119 | Quartz.CFRunLoopRunInMode(Quartz.kCFRunLoopDefaultMode, 5, False)
120 |
121 | def handler(self, proxy, type, event, refcon):
122 | (x, y) = Quartz.CGEventGetLocation(event)
123 | if type in pressID:
124 | self.click(x, y, pressID.index(type), True)
125 | elif type in releaseID:
126 | self.click(x, y, releaseID.index(type), False)
127 | else:
128 | self.move(x, y)
129 |
130 | if self.capture:
131 | Quartz.CGEventSetType(event, Quartz.kCGEventNull)
132 |
133 | return event
134 |
--------------------------------------------------------------------------------
/linux/tensorbox/pymouse/mir.py:
--------------------------------------------------------------------------------
1 | #Copyright 2013 Paul Barton
2 | #
3 | #This program is free software: you can redistribute it and/or modify
4 | #it under the terms of the GNU General Public License as published by
5 | #the Free Software Foundation, either version 3 of the License, or
6 | #(at your option) any later version.
7 | #
8 | #This program is distributed in the hope that it will be useful,
9 | #but WITHOUT ANY WARRANTY; without even the implied warranty of
10 | #MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 | #GNU General Public License for more details.
12 | #
13 | #You should have received a copy of the GNU General Public License
14 | #along with this program. If not, see .
15 |
--------------------------------------------------------------------------------
/linux/tensorbox/pymouse/wayland.py:
--------------------------------------------------------------------------------
1 | #Copyright 2013 Paul Barton
2 | #
3 | #This program is free software: you can redistribute it and/or modify
4 | #it under the terms of the GNU General Public License as published by
5 | #the Free Software Foundation, either version 3 of the License, or
6 | #(at your option) any later version.
7 | #
8 | #This program is distributed in the hope that it will be useful,
9 | #but WITHOUT ANY WARRANTY; without even the implied warranty of
10 | #MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 | #GNU General Public License for more details.
12 | #
13 | #You should have received a copy of the GNU General Public License
14 | #along with this program. If not, see .
15 |
--------------------------------------------------------------------------------
/linux/tensorbox/pymouse/windows.py:
--------------------------------------------------------------------------------
1 | #Copyright 2013 Paul Barton
2 | #
3 | #This program is free software: you can redistribute it and/or modify
4 | #it under the terms of the GNU General Public License as published by
5 | #the Free Software Foundation, either version 3 of the License, or
6 | #(at your option) any later version.
7 | #
8 | #This program is distributed in the hope that it will be useful,
9 | #but WITHOUT ANY WARRANTY; without even the implied warranty of
10 | #MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 | #GNU General Public License for more details.
12 | #
13 | #You should have received a copy of the GNU General Public License
14 | #along with this program. If not, see .
15 |
16 | from ctypes import *
17 | import win32api
18 | import win32con
19 | from .base import PyMouseMeta, PyMouseEventMeta, ScrollSupportError
20 | import pythoncom
21 | from time import sleep
22 |
23 | class POINT(Structure):
24 | _fields_ = [("x", c_ulong),
25 | ("y", c_ulong)]
26 |
27 | class PyMouse(PyMouseMeta):
28 | """MOUSEEVENTF_(button and action) constants
29 | are defined at win32con, buttonAction is that value"""
30 |
31 | def press(self, x, y, button=1):
32 | buttonAction = 2 ** ((2 * button) - 1)
33 | self.move(x, y)
34 | win32api.mouse_event(buttonAction, x, y)
35 |
36 | def release(self, x, y, button=1):
37 | buttonAction = 2 ** ((2 * button))
38 | self.move(x, y)
39 | win32api.mouse_event(buttonAction, x, y)
40 |
41 | def scroll(self, vertical=None, horizontal=None, depth=None):
42 |
43 | #Windows supports only vertical and horizontal scrolling
44 | if depth is not None:
45 | raise ScrollSupportError('PyMouse cannot support depth-scrolling \
46 | in Windows. This feature is only available on Mac.')
47 |
48 | #Execute vertical then horizontal scrolling events
49 | if vertical is not None:
50 | vertical = int(vertical)
51 | if vertical == 0: # Do nothing with 0 distance
52 | pass
53 | elif vertical > 0: # Scroll up if positive
54 | for _ in range(vertical):
55 | win32api.mouse_event(0x0800, 0, 0, 120, 0)
56 | else: # Scroll down if negative
57 | for _ in range(abs(vertical)):
58 | win32api.mouse_event(0x0800, 0, 0, -120, 0)
59 | if horizontal is not None:
60 | horizontal = int(horizontal)
61 | if horizontal == 0: # Do nothing with 0 distance
62 | pass
63 | elif horizontal > 0: # Scroll right if positive
64 | for _ in range(horizontal):
65 | win32api.mouse_event(0x01000, 0, 0, 120, 0)
66 | else: # Scroll left if negative
67 | for _ in range(abs(horizontal)):
68 | win32api.mouse_event(0x01000, 0, 0, -120, 0)
69 |
70 | def move(self, x, y):
71 | windll.user32.SetCursorPos(x, y)
72 |
73 | def drag(self, x, y):
74 | self.press(*self.position())
75 | #self.move(x, y)
76 | self.release(x, y)
77 |
78 | def position(self):
79 | pt = POINT()
80 | windll.user32.GetCursorPos(byref(pt))
81 | return pt.x, pt.y
82 |
83 | def screen_size(self):
84 | if windll.user32.GetSystemMetrics(80) == 1:
85 | width = windll.user32.GetSystemMetrics(0)
86 | height = windll.user32.GetSystemMetrics(1)
87 | else:
88 | width = windll.user32.GetSystemMetrics(78)
89 | height = windll.user32.GetSystemMetrics(79)
90 | return width, height
91 |
92 | class PyMouseEvent(PyMouseEventMeta):
93 | def __init__(self, capture=False, capture_move=False):
94 | import pyHook
95 |
96 | PyMouseEventMeta.__init__(self, capture=capture, capture_move=capture_move)
97 | self.hm = pyHook.HookManager()
98 |
99 | def run(self):
100 | self.hm.MouseAll = self._action
101 | self.hm.HookMouse()
102 | while self.state:
103 | sleep(0.01)
104 | pythoncom.PumpWaitingMessages()
105 |
106 | def stop(self):
107 | self.hm.UnhookMouse()
108 | self.state = False
109 |
110 | def _action(self, event):
111 | import pyHook
112 | x, y = event.Position
113 |
114 | if event.Message == pyHook.HookConstants.WM_MOUSEMOVE:
115 | self.move(x,y)
116 | return not self.capture_move
117 |
118 | elif event.Message == pyHook.HookConstants.WM_LBUTTONDOWN:
119 | self.click(x, y, 1, True)
120 | elif event.Message == pyHook.HookConstants.WM_LBUTTONUP:
121 | self.click(x, y, 1, False)
122 | elif event.Message == pyHook.HookConstants.WM_RBUTTONDOWN:
123 | self.click(x, y, 2, True)
124 | elif event.Message == pyHook.HookConstants.WM_RBUTTONUP:
125 | self.click(x, y, 2, False)
126 | elif event.Message == pyHook.HookConstants.WM_MBUTTONDOWN:
127 | self.click(x, y, 3, True)
128 | elif event.Message == pyHook.HookConstants.WM_MBUTTONUP:
129 | self.click(x, y, 3, False)
130 |
131 | elif event.Message == pyHook.HookConstants.WM_MOUSEWHEEL:
132 | # event.Wheel is -1 when scrolling down, 1 when scrolling up
133 | self.scroll(x, y, event.Wheel, 0)
134 |
135 | return not self.capture
136 |
--------------------------------------------------------------------------------
/linux/tensorbox/pymouse/x11.py:
--------------------------------------------------------------------------------
1 | #Copyright 2013 Paul Barton
2 | #
3 | #This program is free software: you can redistribute it and/or modify
4 | #it under the terms of the GNU General Public License as published by
5 | #the Free Software Foundation, either version 3 of the License, or
6 | #(at your option) any later version.
7 | #
8 | #This program is distributed in the hope that it will be useful,
9 | #but WITHOUT ANY WARRANTY; without even the implied warranty of
10 | #MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 | #GNU General Public License for more details.
12 | #
13 | #You should have received a copy of the GNU General Public License
14 | #along with this program. If not, see .
15 |
16 | from Xlib.display import Display
17 | from Xlib import X
18 | from Xlib.ext.xtest import fake_input
19 | from Xlib.ext import record
20 | from Xlib.protocol import rq
21 |
22 | from .base import PyMouseMeta, PyMouseEventMeta, ScrollSupportError
23 |
24 |
25 | class X11Error(Exception):
26 | """An error that is thrown at the end of a code block managed by a
27 | :func:`display_manager` if an *X11* error occurred.
28 | """
29 | pass
30 |
31 |
32 | def display_manager(display):
33 | """Traps *X* errors and raises an :class:``X11Error`` at the end if any
34 | error occurred.
35 |
36 | This handler also ensures that the :class:`Xlib.display.Display` being
37 | managed is sync'd.
38 |
39 | :param Xlib.display.Display display: The *X* display.
40 |
41 | :return: the display
42 | :rtype: Xlib.display.Display
43 | """
44 | from contextlib import contextmanager
45 |
46 | @contextmanager
47 | def manager():
48 | errors = []
49 |
50 | def handler(*args):
51 | errors.append(args)
52 |
53 | old_handler = display.set_error_handler(handler)
54 | yield display
55 | display.sync()
56 | display.set_error_handler(old_handler)
57 | if errors:
58 | raise X11Error(errors)
59 |
60 | return manager()
61 |
62 |
63 | def translate_button_code(button):
64 | # In X11, the button numbers are:
65 | # leftclick=1, middleclick=2, rightclick=3
66 | # For the purposes of the cross-platform interface of PyMouse, we
67 | # invert the button number values of the right and middle buttons
68 | if button in [1, 2, 3]:
69 | return (None, 1, 3, 2)[button]
70 | else:
71 | return button
72 |
73 | def button_code_to_scroll_direction(button):
74 | # scrollup=4, scrolldown=5, scrollleft=6, scrollright=7
75 | return {
76 | 4: (1, 0),
77 | 5: (-1, 0),
78 | 6: (0, 1),
79 | 7: (0, -1),
80 | }[button]
81 |
82 |
83 | class PyMouse(PyMouseMeta):
84 | def __init__(self, display=None):
85 | PyMouseMeta.__init__(self)
86 | self.display = Display(display)
87 | self.display2 = Display(display)
88 |
89 | def press(self, button=1):
90 | #self.move(x, y)
91 |
92 | with display_manager(self.display) as d:
93 | fake_input(d, X.ButtonPress, translate_button_code(button))
94 |
95 | def release(self, button=1):
96 | #self.move(x, y)
97 |
98 | with display_manager(self.display) as d:
99 | fake_input(d, X.ButtonRelease, translate_button_code(button))
100 |
101 | def scroll(self, vertical=None, horizontal=None, depth=None):
102 | #Xlib supports only vertical and horizontal scrolling
103 | if depth is not None:
104 | raise ScrollSupportError('PyMouse cannot support depth-scrolling \
105 | in X11. This feature is only available on Mac.')
106 |
107 | #Execute vertical then horizontal scrolling events
108 | if vertical is not None:
109 | vertical = int(vertical)
110 | if vertical == 0: # Do nothing with 0 distance
111 | pass
112 | elif vertical > 0: # Scroll up if positive
113 | self.click(*self.position(), button=4, n=vertical)
114 | else: # Scroll down if negative
115 | self.click(*self.position(), button=5, n=abs(vertical))
116 | if horizontal is not None:
117 | horizontal = int(horizontal)
118 | if horizontal == 0: # Do nothing with 0 distance
119 | pass
120 | elif horizontal > 0: # Scroll right if positive
121 | self.click(*self.position(), button=7, n=horizontal)
122 | else: # Scroll left if negative
123 | self.click(*self.position(), button=6, n=abs(horizontal))
124 |
125 | def move(self, x, y):
126 | if (x, y) != self.position():
127 | with display_manager(self.display) as d:
128 | fake_input(d, X.MotionNotify, x=x, y=y)
129 |
130 | def drag(self, x, y):
131 | with display_manager(self.display) as d:
132 | fake_input(d, X.ButtonPress, 1)
133 | fake_input(d, X.MotionNotify, x=x, y=y)
134 | fake_input(d, X.ButtonRelease, 1)
135 |
136 | def position(self):
137 | coord = self.display.screen().root.query_pointer()._data
138 | return coord["root_x"], coord["root_y"]
139 |
140 | def screen_size(self):
141 | width = self.display.screen().width_in_pixels
142 | height = self.display.screen().height_in_pixels
143 | return width, height
144 |
145 |
146 | class PyMouseEvent(PyMouseEventMeta):
147 | def __init__(self, capture=False, capture_move=False, display=None):
148 | PyMouseEventMeta.__init__(self,
149 | capture=capture,
150 | capture_move=capture_move)
151 | self.display = Display(display)
152 | self.display2 = Display(display)
153 | self.ctx = self.display2.record_create_context(
154 | 0,
155 | [record.AllClients],
156 | [{
157 | 'core_requests': (0, 0),
158 | 'core_replies': (0, 0),
159 | 'ext_requests': (0, 0, 0, 0),
160 | 'ext_replies': (0, 0, 0, 0),
161 | 'delivered_events': (0, 0),
162 | 'device_events': (X.ButtonPressMask, X.ButtonReleaseMask),
163 | 'errors': (0, 0),
164 | 'client_started': False,
165 | 'client_died': False,
166 | }])
167 |
168 | def run(self):
169 | try:
170 | if self.capture and self.capture_move:
171 | capturing = X.ButtonPressMask | X.ButtonReleaseMask | X.PointerMotionMask
172 | elif self.capture:
173 | capturing = X.ButtonPressMask | X.ButtonReleaseMask
174 | elif self.capture_move:
175 | capturing = X.PointerMotionMask
176 | else:
177 | capturing = False
178 |
179 | if capturing:
180 | self.display2.screen().root.grab_pointer(True,
181 | capturing,
182 | X.GrabModeAsync,
183 | X.GrabModeAsync,
184 | 0, 0, X.CurrentTime)
185 | self.display.screen().root.grab_pointer(True,
186 | capturing,
187 | X.GrabModeAsync,
188 | X.GrabModeAsync,
189 | 0, 0, X.CurrentTime)
190 |
191 | self.display2.record_enable_context(self.ctx, self.handler)
192 | self.display2.record_free_context(self.ctx)
193 | except KeyboardInterrupt:
194 | self.stop()
195 |
196 | def stop(self):
197 | self.state = False
198 | with display_manager(self.display) as d:
199 | d.ungrab_pointer(X.CurrentTime)
200 | d.record_disable_context(self.ctx)
201 | with display_manager(self.display2) as d:
202 | d.ungrab_pointer(X.CurrentTime)
203 | d.record_disable_context(self.ctx)
204 |
205 | def handler(self, reply):
206 | data = reply.data
207 | while len(data):
208 | event, data = rq.EventField(None).parse_binary_value(data, self.display.display, None, None)
209 |
210 | if event.detail in [4, 5, 6, 7]:
211 | if event.type == X.ButtonPress:
212 | self.scroll(event.root_x, event.root_y, *button_code_to_scroll_direction(event.detail))
213 | elif event.type == X.ButtonPress:
214 | self.click(event.root_x, event.root_y, translate_button_code(event.detail), True)
215 | elif event.type == X.ButtonRelease:
216 | self.click(event.root_x, event.root_y, translate_button_code(event.detail), False)
217 | else:
218 | self.move(event.root_x, event.root_y)
219 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/Makefile:
--------------------------------------------------------------------------------
1 | SHELL := /bin/bash
2 |
3 | .PHONY: all
4 | all:
5 | pip install runcython
6 | makecython++ stitch_wrapper.pyx "" "stitch_rects.cpp ./hungarian/hungarian.cpp"
7 |
8 | hungarian: hungarian/hungarian.so
9 |
10 | hungarian/hungarian.so:
11 | cd hungarian && \
12 | TF_INC=$$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_include())') && \
13 | if [ `uname` == Darwin ];\
14 | then g++ -std=c++11 -shared hungarian.cc -o hungarian.so -fPIC -I -D_GLIBCXX_USE_CXX11_ABI=0$$TF_INC;\
15 | else g++ -std=c++11 -shared hungarian.cc -o hungarian.so -fPIC -I $$TF_INC; fi
16 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/__init__.py:
--------------------------------------------------------------------------------
1 |
2 | import tensorflow as tf
3 | from distutils.version import LooseVersion
4 |
5 | TENSORFLOW_VERSION = LooseVersion(tf.__version__)
6 |
7 | def tf_concat(axis, values, **kwargs):
8 | if TENSORFLOW_VERSION >= LooseVersion('1.0'):
9 | return tf.concat(values, axis, **kwargs)
10 | else:
11 | return tf.concat(axis, values, **kwargs)
12 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/annolist/LICENSE_FOR_THIS_FOLDER:
--------------------------------------------------------------------------------
1 | MPII Human Pose Dataset, Version 1.0
2 | Copyright 2015 Max Planck Institute for Informatics
3 | Licensed under the Simplified BSD License
4 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/annolist/MatPlotter.py:
--------------------------------------------------------------------------------
1 | import os
2 | import sys
3 | import string
4 | import matplotlib
5 | matplotlib.use('Agg')
6 | from pylab import *
7 | import numpy as np
8 |
9 | class MatPlotter:
10 | fontsize=15
11 | color=0
12 | colors=["r-", "b-", "k-", "c-", "m-", "y-"]
13 | colors+=[x + "-" for x in colors]
14 | colors+=["g-", "g--"]
15 | curFigure=[]
16 | legendNames=[]
17 | fontsizeLegend=14
18 | legendPlace='lower right'
19 | legendborderpad = None
20 | legendlabelsep = None
21 |
22 |
23 | def __init__(self, fontsize=15):
24 | # self.newFigure()
25 | self.fontsize=fontsize
26 | self.fontsizeLegend=fontsize - 1
27 | pass
28 |
29 | def formatLegend(self, newFontSize = 14, newPlace = 'lower right', borderpad = None, labelsep = None):
30 | self.fontsizeLegend=newFontSize
31 | self.legendPlace=newPlace
32 | self.legendborderpad = borderpad
33 | self.legendlabelsep = labelsep
34 |
35 | def newFigure(self, plotTitle="", fsize=rcParams['figure.figsize']):
36 | return self.newRPCFigure(plotTitle, fsize)
37 |
38 | def newRPCFigure(self, plotTitle="", fsize=rcParams['figure.figsize']):
39 | curFigure = figure(figsize=fsize)
40 | self.title = title(plotTitle, fontsize=self.fontsize)
41 | #subplots_adjust(left=0.085, right=0.975, top=0.975, bottom=0.085)
42 | subplots_adjust(right=0.975, top=0.975)
43 |
44 | #axis('equal')
45 | axis([0, 1, 0, 1])
46 | xticklocs, xticklabels = xticks(arange(0, 1.01, 0.1))
47 | setp(xticklabels, size=self.fontsize)
48 | yticklocs, yticklabels = yticks(arange(0, 1.01, 0.1))
49 | setp(yticklabels, size=self.fontsize)
50 | self.xlabel = xlabel("1-precision")
51 | self.xlabel.set_size(self.fontsize+2)
52 | self.ylabel = ylabel("recall")
53 | self.ylabel.set_size(self.fontsize+4)
54 | grid()
55 | hold(True)
56 |
57 | def newFPPIFigure(self, plotTitle="", fsize=rcParams['figure.figsize']):
58 | curFigure = figure(figsize=fsize)
59 | self.title = title(plotTitle, fontsize=self.fontsize)
60 | subplots_adjust(left=0.085, right=0.975, top=0.975, bottom=0.085)
61 |
62 | #axis('equal')
63 | axis([0, 100, 0, 1])
64 | xticklocs, xticklabels = xticks(arange(0, 100.01, 0.5))
65 | setp(xticklabels, size=self.fontsize)
66 | yticklocs, yticklabels = yticks(arange(0, 1.01, 0.1))
67 | setp(yticklabels, size=self.fontsize)
68 | self.xlabel = xlabel("false positives per image")
69 | self.xlabel.set_size(self.fontsize+2)
70 | self.ylabel = ylabel("recall")
71 | self.ylabel.set_size(self.fontsize+4)
72 | grid()
73 | hold(True)
74 |
75 |
76 | def newFreqFigure(self, plotTitle="", maxX = 10, maxY = 10,fsize=rcParams['figure.figsize']):
77 | curFigure = figure(figsize=fsize)
78 | self.title = title(plotTitle, fontsize=self.fontsize)
79 | subplots_adjust(left=0.085, right=0.975, top=0.975, bottom=0.1)
80 | #axis('equal')
81 |
82 | axis([0, maxX, 0, maxY])
83 | xticklocs, xticklabels = xticks(arange(0, maxX + 0.01, maxX * 1.0/ 10))
84 | setp(xticklabels, size=self.fontsize)
85 | yticklocs, yticklabels = yticks(arange(0, maxY + 0.01, maxY * 1.0/ 10))
86 | setp(yticklabels, size=self.fontsize)
87 | self.xlabel = xlabel("False positive / ground truth rect")
88 | self.xlabel.set_size(self.fontsize+2)
89 | self.ylabel = ylabel("True positives / ground truth rect")
90 | self.ylabel.set_size(self.fontsize+4)
91 | grid()
92 | hold(True)
93 |
94 | def newFPPWFigure(self, plotTitle="", fsize=rcParams['figure.figsize']):
95 | curFigure = figure(figsize=fsize)
96 | self.title = title(plotTitle, fontsize=self.fontsize)
97 | subplots_adjust(left=0.085, right=0.975, top=0.975, bottom=0.085)
98 |
99 | self.xlabel = xlabel("false positive per windows (FPPW)")
100 | self.xlabel.set_size(self.fontsize+2)
101 | self.ylabel = ylabel("miss rate")
102 | self.ylabel.set_size(self.fontsize+4)
103 |
104 | grid()
105 | hold(True)
106 |
107 | def newLogFPPIFigure(self, plotTitle="", fsize=rcParams['figure.figsize']):
108 | curFigure = figure(figsize=fsize)
109 | self.title = title(plotTitle, fontsize=self.fontsize)
110 | subplots_adjust(left=0.085, right=0.975, top=0.975, bottom=0.1)
111 |
112 | #axis('equal')
113 |
114 | self.xlabel = xlabel("false positives per image")
115 | self.xlabel.set_size(self.fontsize+2)
116 | self.ylabel = ylabel("miss rate")
117 | self.ylabel.set_size(self.fontsize+4)
118 | grid()
119 | hold(True)
120 |
121 | def loadRPCData(self, fname):
122 | self.filename = fname
123 | self.prec=[]
124 | self.rec=[]
125 | self.score=[]
126 | self.fppi=[]
127 | file = open(fname)
128 |
129 | precScores = []
130 | for i in range(1,10,1):
131 | precScores.append(100 - i * 10)
132 |
133 | fppiScores=[]
134 | for i in range(0, 500, 5):
135 | fppiScores.append(i * 1.0 / 100.0)
136 |
137 |
138 |
139 | precinfo = []
140 | fppiinfo = []
141 | eerinfo = []
142 | logAvInfo = []
143 |
144 | logAvMR= []
145 | self.lamr = 0;
146 | self.eer = None;
147 | firstLine = True
148 | leadingZeroCount = 0
149 |
150 | for line in file.readlines():
151 | vals = line.split()
152 | #vals=line.split(" ")
153 | #for val in vals:
154 | # if val=="":
155 | # vals.remove(val)
156 | self.prec.append(1-float(vals[0]))
157 | self.rec.append(float(vals[1]))
158 | self.score.append(float(vals[2]))
159 |
160 | if(len(vals)>3):
161 | self.fppi.append(float(vals[3]))
162 | if firstLine and not float(vals[3]) == 0:
163 | firstLine = False
164 |
165 | lamrcount = 1
166 | self.lamr = 1 - float(vals[1])
167 |
168 | lowest_fppi = math.ceil( math.log(float(vals[3]))/ math.log(10) * 10 )
169 | print "lowest_fppi: ",lowest_fppi;
170 |
171 | # MA: temporarily commented out
172 | # for i in range(lowest_fppi, 1, 1):
173 | # logAvMR.append(10** (i * 1.0 / 10))
174 |
175 | #self.score.append(float(vals[2][:-1]))
176 | #print 1-self.prec[-1], self.rec[-1], self.score[-1]
177 | if (len(self.prec)>1):
178 | diff = (1-self.prec[-1]-self.rec[-1]) * (1-self.prec[-2]-self.rec[-2])
179 | if ( diff <0):
180 | eerinfo.append( "EER between: %.03f and %.03f\tScore:%f" % (self.rec[-1], self.rec[-2], self.score[-1]))
181 | self.eer = (self.rec[-1]+self.rec[-2]) * 0.5
182 | if ( diff == 0 and 1-self.prec[-1]-self.rec[-1]==0):
183 | eerinfo.append( "EER: %.03f\tScore:%f" % (self.rec[-1], self.score[-1]))
184 | self.eer = self.rec[-1]
185 |
186 | #Remove already passed precision
187 | if (len(precScores) > 0 and (float(vals[0])) < precScores[0] / 100.0):
188 | precinfo.append("%d percent precision score: %f, recall: %.03f" % (precScores[0], float(vals[2]), float(vals[1])))
189 | while(len(precScores) > 0 and precScores[0]/100.0 > float(vals[0])):
190 | precScores.pop(0)
191 |
192 | #Remove already passed precision
193 | if(len(vals) > 3):
194 | if (len(fppiScores) > 0 and (float(vals[3])) > fppiScores[0]):
195 | fppiinfo.append("%f fppi score: %f, recall: %.03f" % (fppiScores[0], float(vals[2]), float(vals[1])))
196 | while(len(fppiScores) > 0 and fppiScores[0] < float(vals[3])):
197 | fppiScores.pop(0)
198 |
199 | if (len(logAvMR) > 0 and (float(vals[3])) > logAvMR[0]):
200 | while(len(logAvMR) > 0 and logAvMR[0] < float(vals[3])):
201 | logAvInfo.append("%f fppi, miss rate: %.03f, score: %f" % (logAvMR[0], 1-float(vals[1]), float(vals[2])) )
202 | self.lamr += 1-float(vals[1])
203 | lamrcount += 1
204 | logAvMR.pop(0)
205 |
206 | lastMR = 1-float(vals[1])
207 |
208 |
209 | if(len(vals)>3):
210 | for i in logAvMR:
211 | logAvInfo.append("%f fppi, miss rate: %.03f, extended" % (i, lastMR) )
212 | self.lamr += lastMR
213 | lamrcount += 1
214 |
215 | for i in precinfo:
216 | print i;
217 | print;
218 | for i in fppiinfo:
219 | print i;
220 | print
221 | for i in eerinfo:
222 | print i;
223 | print
224 | print "Recall at first false positive: %.03f" % self.rec[0]
225 | if(len(vals)>3):
226 | print
227 | for i in logAvInfo:
228 | print i;
229 | self.lamr = self.lamr * 1.0 / lamrcount
230 | print "Log average miss rate in [10^%.01f, 10^0]: %.03f" % (lowest_fppi / 10.0, self.lamr )
231 |
232 |
233 |
234 | print; print
235 | file.close()
236 |
237 | def loadFreqData(self, fname):
238 | self.filename = fname
239 | self.prec=[]
240 | self.rec=[]
241 | self.score=[]
242 | file = open(fname)
243 |
244 | for line in file.readlines():
245 | vals = line.split()
246 |
247 | self.prec.append(float(vals[0]))
248 | self.rec.append(float(vals[1]))
249 | self.score.append(float(vals[2]))
250 |
251 | file.close()
252 |
253 | def loadFPPWData(self, fname):
254 | self.loadFreqData(fname)
255 |
256 | def finishPlot(self, axlimits = [0,1.0,0,1.0]):
257 | # MA:
258 | #self.legend = legend(self.legendNames, self.legendPlace, pad = self.legendborderpad, labelsep = self.legendlabelsep)
259 | self.legend = legend(self.legendNames, self.legendPlace)
260 |
261 | lstrings = self.legend.get_texts()
262 | setp(lstrings, fontsize=self.fontsizeLegend)
263 | #line= plot( [1 - axlimits[0], 0], [axlimits[3], 1 - axlimits[3] ] , 'k')
264 | line= plot( [1, 0], [0, 1] , 'k')
265 |
266 | def finishFreqPlot(self):
267 | self.legend = legend(self.legendNames, self.legendPlace, pad = self.legendborderpad, labelsep = self.legendlabelsep)
268 | lstrings = self.legend.get_texts()
269 | setp(lstrings, fontsize=self.fontsizeLegend)
270 |
271 |
272 | def show(self, plotEER = True, axlimits = [0,1.0,0,1.0]):
273 | if (plotEER):
274 | self.finishPlot(axlimits)
275 | axis(axlimits)
276 | else:
277 | self.finishFreqPlot()
278 |
279 | show()
280 |
281 | def saveCurrentFigure(self, plotEER, filename, axlimits = [0,1.0,0,1.0]):
282 | if (plotEER):
283 | self.finishPlot(axlimits)
284 | axis(axlimits)
285 | else:
286 | self.finishFreqPlot()
287 |
288 | print "Saving: " + filename
289 | savefig(filename)
290 |
291 | def plotRFP(self, numImages, fname, line="r-"):
292 | print 'NOT YET IMPLEMENTED'
293 |
294 | def plotRPC(self, fname, descr="line", style="-1", axlimits = [0,1.0,0,1.0], linewidth = 2, dashstyle = [], addEER = False ):
295 | self.loadRPCData(fname)
296 |
297 | #axis(axlimits);
298 | if (style=="-1"):
299 | if dashstyle != []:
300 | line = plot(self.prec, self.rec, self.colors[self.color], dashes = dashstyle)
301 | else:
302 | line = plot(self.prec, self.rec, self.colors[self.color])
303 | self.color=self.color+1
304 | self.color=self.color % len(self.colors)
305 | else:
306 | if dashstyle != []:
307 | line = plot(self.prec, self.rec, style, dashes = dashstyle)
308 | else:
309 | line = plot(self.prec, self.rec, style)
310 |
311 | axis(axlimits)
312 |
313 | if addEER and self.eer != None:
314 | descr += " (%.01f%%)" % (self.eer * 100)
315 |
316 | setp(line, 'linewidth', linewidth)
317 | self.legendNames= self.legendNames+[descr]
318 |
319 | def plotFPPI(self, fname, descr="line", style="-1", axlimits = [0,2,0,1], linewidth = 2, dashstyle = []):
320 | self.loadRPCData(fname)
321 |
322 | if (style=="-1"):
323 | if dashstyle != []:
324 | line = plot(self.fppi, self.rec, self.colors[self.color], dashes = dashstyle)
325 | else:
326 | line = plot(self.fppi, self.rec, self.colors[self.color])
327 | self.color=self.color+1
328 | self.color=self.color % len(self.colors)
329 | else:
330 | if dashstyle != []:
331 | line = plot(self.fppi, self.rec, style, dashes = dashstyle)
332 | else:
333 | line = plot(self.fppi, self.rec, style)
334 |
335 | axis(axlimits);
336 |
337 | setp(line, 'linewidth', linewidth)
338 | self.legendNames= self.legendNames+[descr]
339 |
340 |
341 | def plotFreq(self, fname, descr="line", style="-1", linewidth = 2, dashstyle = []):
342 | self.loadFreqData(fname)
343 | if (style=="-1"):
344 | if dashstyle != []:
345 | line = plot(self.prec, self.rec, self.colors[self.color], dashes = dashstyle)
346 | else:
347 | line = plot(self.prec, self.rec, self.colors[self.color])
348 | self.color=self.color+1
349 | self.color=self.color % len(self.colors)
350 | else:
351 | if dashstyle != []:
352 | line = plot(self.prec, self.rec, style, dashes = dashstyle)
353 | else:
354 | line = plot(self.prec, self.rec, style)
355 |
356 |
357 | setp(line, 'linewidth', linewidth)
358 | self.legendNames= self.legendNames+[descr]
359 |
360 | def plotFPPW(self, fname, descr="line", style="-1", axlimits = [5e-6, 1e0, 1e-2, 0.5], linewidth = 2, dashstyle = []):
361 | self.loadFPPWData(fname)
362 | if (style=="-1"):
363 | if dashstyle != []:
364 | line = loglog(self.prec, self.rec, self.colors[self.color], dashes = dashstyle)
365 | else:
366 | line = loglog(self.prec, self.rec, self.colors[self.color])
367 | self.color=self.color+1
368 | self.color=self.color % len(self.colors)
369 | else:
370 | if dashstyle != []:
371 | line = loglog(self.prec, self.rec, style, dashes = dashstyle)
372 | else:
373 | line = loglog(self.prec, self.rec, style)
374 |
375 | xticklocs, xticklabels = xticks([1e-5, 1e-4,1e-3, 1e-2, 1e-1, 1e0])
376 | setp(xticklabels, size=self.fontsize)
377 | yticklocs, yticklabels = yticks(array([0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09, 0.1, 0.2, 0.3, 0.4, 0.5]),
378 | ("0.01", "0.02", "0.03", "0.04", "0.05", "0.06", "0.07", "0.08","0.09", "0.1", "0.2", "0.3", "0.4", "0.5"))
379 | setp(yticklabels, size=self.fontsize)
380 |
381 | axis(axlimits)
382 |
383 | gca().yaxis.grid(True, 'minor')
384 | setp(line, 'linewidth', linewidth)
385 |
386 | self.legendNames= self.legendNames+[descr]
387 |
388 | def plotLogFPPI(self, fname, descr="line", style="-1", axlimits = [5e-3, 1e1, 1e-1, 1], linewidth = 2, dashstyle = [], addlamr = False):
389 | self.loadRPCData(fname)
390 | if (style=="-1"):
391 | if dashstyle != []:
392 | line = loglog(self.fppi, [1 - x for x in self.rec], self.colors[self.color], dashes = dashstyle)
393 | else:
394 | line = loglog(self.fppi, [1 - x for x in self.rec], self.colors[self.color])
395 |
396 | self.color=(self.color+1) % len(self.colors)
397 | else:
398 | if dashstyle != []:
399 | line = loglog(self.fppi, [1 - x for x in self.rec], style, dashes = dashstyle)
400 | else:
401 | line = loglog(self.fppi, [1 - x for x in self.rec], style)
402 |
403 | gca().yaxis.grid(True, 'minor')
404 |
405 | m = min(self.fppi)
406 | lax = axlimits[0]
407 | for i in self.fppi:
408 | if(i != m):
409 | lax = math.floor(log(i)/math.log(10))
410 | leftlabel = math.pow(10, lax)
411 | break
412 |
413 | m = max(self.fppi)
414 | rightlabel = math.pow(10, math.ceil(log(m)/math.log(10))) + 0.01
415 |
416 | k = leftlabel
417 | ticks = [k]
418 | while k < rightlabel:
419 | k = k * 10
420 | ticks.append(k)
421 |
422 | xticklocs, xticklabels = xticks(ticks)
423 | setp(xticklabels, size=self.fontsize)
424 | yticklocs, yticklabels = yticks(arange(0.1, 1.01, 0.1), ("0.1", "0.2", "0.3", "0.4", "0.5", "0.6", "0.7", "0.8", "0.9", "1.0"))
425 | setp(yticklabels, size=self.fontsize)
426 |
427 | axlimits[0] = lax
428 | axis(axlimits)
429 |
430 | setp(line, 'linewidth', linewidth)
431 |
432 | if addlamr:
433 | descr += " (%.01f%%)" % (self.lamr * 100)
434 |
435 | self.legendNames= self.legendNames+[descr]
436 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/annolist/PalLib.py:
--------------------------------------------------------------------------------
1 | import sys
2 | #import AnnoList_pb2
3 | import AnnotationLib;
4 |
5 | from ma_utils import is_number;
6 |
7 | def loadPal(filename):
8 | _annolist = AnnoList_pb2.AnnoList();
9 |
10 | f = open(filename, "rb");
11 | _annolist.ParseFromString(f.read());
12 | f.close();
13 |
14 | return _annolist;
15 |
16 | def savePal(filename, _annolist):
17 | f = open(filename, "wb");
18 | f.write(_annolist.SerializeToString());
19 | f.close();
20 |
21 | def al2pal(annotations):
22 | _annolist = AnnoList_pb2.AnnoList();
23 |
24 | #assert(isinstance(annotations, AnnotationLib.AnnoList));
25 |
26 | # check type of attributes, add missing attributes
27 | for a in annotations:
28 | for r in a.rects:
29 | for k, v in r.at.iteritems():
30 | if not k in annotations.attribute_desc:
31 | annotations.add_attribute(k, type(v));
32 | else:
33 | assert(AnnotationLib.is_compatible_attr_type(annotations.attribute_desc[k].dtype, type(v)));
34 |
35 | # check attributes values
36 | for a in annotations:
37 | for r in a.rects:
38 | for k, v in r.at.iteritems():
39 | if k in annotations.attribute_val_to_str:
40 | # don't allow undefined values
41 | if not v in annotations.attribute_val_to_str[k]:
42 | print "attribute: {}, undefined value: {}".format(k, v);
43 | assert(False);
44 |
45 | # store attribute descriptions in pal structure
46 | for aname, adesc in annotations.attribute_desc.iteritems():
47 | _annolist.attribute_desc.extend([adesc]);
48 |
49 | for a in annotations:
50 | _a = _annolist.annotation.add();
51 | _a.imageName = a.imageName;
52 |
53 | for r in a.rects:
54 | _r = _a.rect.add();
55 |
56 | _r.x1 = r.x1;
57 | _r.y1 = r.y1;
58 | _r.x2 = r.x2;
59 | _r.y2 = r.y2;
60 |
61 | _r.score = float(r.score);
62 |
63 | if hasattr(r, 'id'):
64 | _r.id = r.id;
65 |
66 | if hasattr(r, 'track_id'):
67 | _r.track_id = r.track_id;
68 |
69 | if hasattr(r, 'at'):
70 | for k, v in r.at.items():
71 | _at = _r.attribute.add();
72 |
73 | _at.id = annotations.attribute_desc[k].id;
74 |
75 | if annotations.attribute_desc[k].dtype == AnnotationLib.AnnoList.TYPE_INT32:
76 | assert(AnnotationLib.is_compatible_attr_type(AnnotationLib.AnnoList.TYPE_INT32, type(v)));
77 | _at.val = int(v);
78 | elif annotations.attribute_desc[k].dtype == AnnotationLib.AnnoList.TYPE_FLOAT:
79 | assert(AnnotationLib.is_compatible_attr_type(AnnotationLib.AnnoList.TYPE_FLOAT, type(v)));
80 | _at.fval = float(v);
81 | elif annotations.attribute_desc[k].dtype == AnnotationLib.AnnoList.TYPE_STRING:
82 | assert(AnnotationLib.is_compatible_attr_type(AnnotationLib.AnnoList.TYPE_STRING, type(v)));
83 | _at.strval = str(v);
84 | else:
85 | assert(false);
86 |
87 | return _annolist;
88 |
89 | def pal2al(_annolist):
90 | #annotations = [];
91 | annotations = AnnotationLib.AnnoList();
92 |
93 | for adesc in _annolist.attribute_desc:
94 | annotations.attribute_desc[adesc.name] = adesc;
95 | print "attribute: ", adesc.name, adesc.id
96 |
97 | for valdesc in adesc.val_to_str:
98 | annotations.add_attribute_val(adesc.name, valdesc.s, valdesc.id);
99 |
100 | attribute_name_from_id = {adesc.id: aname for aname, adesc in annotations.attribute_desc.iteritems()}
101 | attribute_dtype_from_id = {adesc.id: adesc.dtype for aname, adesc in annotations.attribute_desc.iteritems()}
102 |
103 | for _a in _annolist.annotation:
104 | anno = AnnotationLib.Annotation()
105 |
106 | anno.imageName = _a.imageName;
107 |
108 | anno.rects = [];
109 |
110 | for _r in _a.rect:
111 | rect = AnnotationLib.AnnoRect()
112 |
113 | rect.x1 = _r.x1;
114 | rect.x2 = _r.x2;
115 | rect.y1 = _r.y1;
116 | rect.y2 = _r.y2;
117 |
118 | if _r.HasField("id"):
119 | rect.id = _r.id;
120 |
121 | if _r.HasField("track_id"):
122 | rect.track_id = _r.track_id;
123 |
124 | if _r.HasField("score"):
125 | rect.score = _r.score;
126 |
127 | for _at in _r.attribute:
128 | try:
129 | cur_aname = attribute_name_from_id[_at.id];
130 | cur_dtype = attribute_dtype_from_id[_at.id];
131 | except KeyError as e:
132 | print "attribute: ", _at.id
133 | print e
134 | assert(False);
135 |
136 | if cur_dtype == AnnotationLib.AnnoList.TYPE_INT32:
137 | rect.at[cur_aname] = _at.val;
138 | elif cur_dtype == AnnotationLib.AnnoList.TYPE_FLOAT:
139 | rect.at[cur_aname] = _at.fval;
140 | elif cur_dtype == AnnotationLib.AnnoList.TYPE_STRING:
141 | rect.at[cur_aname] = _at.strval;
142 | else:
143 | assert(False);
144 |
145 | anno.rects.append(rect);
146 |
147 | annotations.append(anno);
148 |
149 | return annotations;
150 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/annolist/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/bethesirius/ChosunTruck/889644385ce57f971ec2921f006fbb0a167e6f1e/linux/tensorbox/utils/annolist/__init__.py
--------------------------------------------------------------------------------
/linux/tensorbox/utils/annolist/ma_utils.py:
--------------------------------------------------------------------------------
1 | def is_number(s):
2 | try:
3 | float(s)
4 | return True
5 | except ValueError:
6 | return False
7 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/annolist/plotSimple.py:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env python
2 |
3 | import sys
4 | import os
5 | import random
6 | import re
7 | from AnnotationLib import *
8 | from MatPlotter import *
9 | from optparse import OptionParser
10 | from copy import deepcopy
11 | from math import sqrt
12 |
13 |
14 | def main(argv):
15 | parser = OptionParser(usage="usage: %prog [options] [...]")
16 | parser.add_option("-o", "--output-file", action="store",
17 | dest="output", type="str", help="outfile. mandatory")
18 | parser.add_option("--fppw", action="store_true", dest="fppw", help="False Positives Per Window")
19 | parser.add_option("--colors", action="store", dest="colors", help="colors")
20 | parser.add_option("--fppi", action="store_true", dest="fppi", help="False Positives Per Image")
21 | parser.add_option("--lfppi", action="store_true", dest="lfppi", help="False Positives Per Image(log)")
22 | parser.add_option("-c", "--components", action="store", dest="ncomponents", type="int", help="show n trailing components of the part", default=3)
23 | parser.add_option("--cut-trailing", action="store", dest="cutcomponents", type="int", help="cut n trailing components of the part (applied after --components)", default=-1)
24 | parser.add_option("-t", "--title", action="store", dest="title", type="str", default="")
25 | parser.add_option("-f", "--fontsize", action="store", dest="fontsize", type="int", default=12)
26 | parser.add_option("-l", "--legend'", action="store", dest="legend", type="string", default="lr")
27 | (options, args) = parser.parse_args()
28 | plotter = MatPlotter(options.fontsize)
29 |
30 | position = "lower right"
31 | if(options.legend == "ur"):
32 | position = "upper right"
33 | if(options.legend == "ul"):
34 | position = "upper left"
35 | if(options.legend == "ll"):
36 | position = "lower left"
37 | plotter.formatLegend(options.fontsize, newPlace = position)
38 |
39 | title = options.title
40 | colors = None
41 | if (options.colors):
42 | colors = options.colors.split()
43 | if (options.fppw):
44 | plotter.newFPPWFigure(title)
45 | elif (options.lfppi):
46 | plotter.newLogFPPIFigure(title)
47 | elif (options.fppi):
48 | plotter.newFPPIFigure(title)
49 | else:
50 | plotter.newFigure(title)
51 |
52 | for i, filename in enumerate(args):
53 | if (os.path.isdir(filename)):
54 | filename = os.path.join(filename, "rpc", "result-minh-48")
55 | displayname = filename
56 | if (options.ncomponents > 0):
57 | suffix = None
58 | for idx in xrange(options.ncomponents):
59 | displayname, last = os.path.split(displayname)
60 | if (suffix):
61 | suffix = os.path.join(last, suffix)
62 | else:
63 | suffix = last
64 | displayname = suffix
65 | if (options.cutcomponents > 0):
66 | for idx in xrange(options.cutcomponents):
67 | displayname, last = os.path.split(displayname)
68 | # plusidx = displayname.index("+")
69 | # displayname = displayname[plusidx:]
70 | print "Plotting: "+displayname
71 | if (options.fppw):
72 | plotter.plotFPPW(filename, displayname)
73 | elif (options.lfppi):
74 | if colors:
75 | plotter.plotLogFPPI(filename, displayname, colors[i])
76 | else:
77 | plotter.plotLogFPPI(filename, displayname)
78 | elif (options.fppi):
79 | plotter.plotFPPI(filename, displayname)
80 | else:
81 | plotter.plotRPC(filename, displayname)
82 |
83 | plotLine = not (options.fppw or options.lfppi or options.fppi);
84 |
85 | if (options.output is None):
86 | plotter.show(plotLine)
87 | else:
88 | plotter.saveCurrentFigure(plotLine, options.output)
89 | return 0
90 |
91 | if __name__ == "__main__":
92 | sys.exit(main(sys.argv))
93 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/data_utils.py:
--------------------------------------------------------------------------------
1 | import os
2 | import cv2
3 | import re
4 | import sys
5 | import argparse
6 | import numpy as np
7 | import copy
8 | import annolist.AnnotationLib as al
9 |
10 | def annotation_to_h5(H, a, cell_width, cell_height, max_len):
11 | region_size = H['region_size']
12 | assert H['region_size'] == H['image_height'] / H['grid_height']
13 | assert H['region_size'] == H['image_width'] / H['grid_width']
14 | cell_regions = get_cell_grid(cell_width, cell_height, region_size)
15 |
16 | cells_per_image = len(cell_regions)
17 |
18 | box_list = [[] for idx in range(cells_per_image)]
19 |
20 | for cidx, c in enumerate(cell_regions):
21 | box_list[cidx] = [r for r in a.rects if all(r.intersection(c))]
22 |
23 | boxes = np.zeros((1, cells_per_image, 4, max_len, 1), dtype = np.float)
24 | box_flags = np.zeros((1, cells_per_image, 1, max_len, 1), dtype = np.float)
25 |
26 | for cidx in xrange(cells_per_image):
27 | #assert(cur_num_boxes <= max_len)
28 |
29 | cell_ox = 0.5 * (cell_regions[cidx].x1 + cell_regions[cidx].x2)
30 | cell_oy = 0.5 * (cell_regions[cidx].y1 + cell_regions[cidx].y2)
31 |
32 | unsorted_boxes = []
33 | for bidx in xrange(min(len(box_list[cidx]), max_len)):
34 |
35 | # relative box position with respect to cell
36 | ox = 0.5 * (box_list[cidx][bidx].x1 + box_list[cidx][bidx].x2) - cell_ox
37 | oy = 0.5 * (box_list[cidx][bidx].y1 + box_list[cidx][bidx].y2) - cell_oy
38 |
39 | width = abs(box_list[cidx][bidx].x2 - box_list[cidx][bidx].x1)
40 | height= abs(box_list[cidx][bidx].y2 - box_list[cidx][bidx].y1)
41 |
42 | if (abs(ox) < H['focus_size'] * region_size and abs(oy) < H['focus_size'] * region_size and
43 | width < H['biggest_box_px'] and height < H['biggest_box_px']):
44 | unsorted_boxes.append(np.array([ox, oy, width, height], dtype=np.float))
45 |
46 | for bidx, box in enumerate(sorted(unsorted_boxes, key=lambda x: x[0]**2 + x[1]**2)):
47 | boxes[0, cidx, :, bidx, 0] = box
48 | box_flags[0, cidx, 0, bidx, 0] = max(box_list[cidx][bidx].silhouetteID, 1)
49 |
50 | return boxes, box_flags
51 |
52 | def get_cell_grid(cell_width, cell_height, region_size):
53 |
54 | cell_regions = []
55 | for iy in xrange(cell_height):
56 | for ix in xrange(cell_width):
57 | cidx = iy * cell_width + ix
58 | ox = (ix + 0.5) * region_size
59 | oy = (iy + 0.5) * region_size
60 |
61 | r = al.AnnoRect(ox - 0.5 * region_size, oy - 0.5 * region_size,
62 | ox + 0.5 * region_size, oy + 0.5 * region_size)
63 | r.track_id = cidx
64 |
65 | cell_regions.append(r)
66 |
67 |
68 | return cell_regions
69 |
70 | def annotation_jitter(I, a_in, min_box_width=20, jitter_scale_min=0.9, jitter_scale_max=1.1, jitter_offset=16, target_width=640, target_height=480):
71 | a = copy.deepcopy(a_in)
72 |
73 | # MA: sanity check
74 | new_rects = []
75 | for i in range(len(a.rects)):
76 | r = a.rects[i]
77 | try:
78 | assert(r.x1 < r.x2 and r.y1 < r.y2)
79 | new_rects.append(r)
80 | except:
81 | print('bad rectangle')
82 | a.rects = new_rects
83 |
84 |
85 | if a.rects:
86 | cur_min_box_width = min([r.width() for r in a.rects])
87 | else:
88 | cur_min_box_width = min_box_width / jitter_scale_min
89 |
90 | # don't downscale below min_box_width
91 | jitter_scale_min = max(jitter_scale_min, float(min_box_width) / cur_min_box_width)
92 |
93 | # it's always ok to upscale
94 | jitter_scale_min = min(jitter_scale_min, 1.0)
95 |
96 | jitter_scale_max = jitter_scale_max
97 |
98 | jitter_scale = np.random.uniform(jitter_scale_min, jitter_scale_max)
99 |
100 | jitter_flip = np.random.random_integers(0, 1)
101 |
102 | if jitter_flip == 1:
103 | I = np.fliplr(I)
104 |
105 | for r in a:
106 | r.x1 = I.shape[1] - r.x1
107 | r.x2 = I.shape[1] - r.x2
108 | r.x1, r.x2 = r.x2, r.x1
109 |
110 | for p in r.point:
111 | p.x = I.shape[1] - p.x
112 |
113 | I1 = cv2.resize(I, None, fx=jitter_scale, fy=jitter_scale, interpolation = cv2.INTER_CUBIC)
114 |
115 | jitter_offset_x = np.random.random_integers(-jitter_offset, jitter_offset)
116 | jitter_offset_y = np.random.random_integers(-jitter_offset, jitter_offset)
117 |
118 |
119 |
120 | rescaled_width = I1.shape[1]
121 | rescaled_height = I1.shape[0]
122 |
123 | px = round(0.5*(target_width)) - round(0.5*(rescaled_width)) + jitter_offset_x
124 | py = round(0.5*(target_height)) - round(0.5*(rescaled_height)) + jitter_offset_y
125 |
126 | I2 = np.zeros((target_height, target_width, 3), dtype=I1.dtype)
127 |
128 | x1 = max(0, px)
129 | y1 = max(0, py)
130 | x2 = min(rescaled_width, target_width - x1)
131 | y2 = min(rescaled_height, target_height - y1)
132 |
133 | I2[0:(y2 - y1), 0:(x2 - x1), :] = I1[y1:y2, x1:x2, :]
134 |
135 | ox1 = round(0.5*rescaled_width) + jitter_offset_x
136 | oy1 = round(0.5*rescaled_height) + jitter_offset_y
137 |
138 | ox2 = round(0.5*target_width)
139 | oy2 = round(0.5*target_height)
140 |
141 | for r in a:
142 | r.x1 = round(jitter_scale*r.x1 - x1)
143 | r.x2 = round(jitter_scale*r.x2 - x1)
144 |
145 | r.y1 = round(jitter_scale*r.y1 - y1)
146 | r.y2 = round(jitter_scale*r.y2 - y1)
147 |
148 | if r.x1 < 0:
149 | r.x1 = 0
150 |
151 | if r.y1 < 0:
152 | r.y1 = 0
153 |
154 | if r.x2 >= I2.shape[1]:
155 | r.x2 = I2.shape[1] - 1
156 |
157 | if r.y2 >= I2.shape[0]:
158 | r.y2 = I2.shape[0] - 1
159 |
160 | for p in r.point:
161 | p.x = round(jitter_scale*p.x - x1)
162 | p.y = round(jitter_scale*p.y - y1)
163 |
164 | # MA: make sure all points are inside the image
165 | r.point = [p for p in r.point if p.x >=0 and p.y >=0 and p.x < I2.shape[1] and p.y < I2.shape[0]]
166 |
167 | new_rects = []
168 | for r in a.rects:
169 | if r.x1 <= r.x2 and r.y1 <= r.y2:
170 | new_rects.append(r)
171 | else:
172 | pass
173 |
174 | a.rects = new_rects
175 |
176 | return I2, a
177 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/googlenet_load.py:
--------------------------------------------------------------------------------
1 | from slim_nets import inception_v1 as inception
2 | from slim_nets import resnet_v1 as resnet
3 | import tensorflow.contrib.slim as slim
4 |
5 | def model(x, H, reuse, is_training=True):
6 | if H['slim_basename'] == 'resnet_v1_101':
7 | with slim.arg_scope(resnet.resnet_arg_scope()):
8 | _, T = resnet.resnet_v1_101(x,
9 | is_training=is_training,
10 | num_classes=1000,
11 | reuse=reuse)
12 | elif H['slim_basename'] == 'InceptionV1':
13 | with slim.arg_scope(inception.inception_v1_arg_scope()):
14 | _, T = inception.inception_v1(x,
15 | is_training=is_training,
16 | num_classes=1001,
17 | spatial_squeeze=False,
18 | reuse=reuse)
19 | #print '\n'.join(map(str, [(k, v.op.outputs[0].get_shape()) for k, v in T.iteritems()]))
20 |
21 | coarse_feat = T[H['slim_top_lname']][:, :, :, :H['later_feat_channels']]
22 | assert coarse_feat.op.outputs[0].get_shape()[3] == H['later_feat_channels']
23 |
24 | # fine feat can be used to reinspect input
25 | attention_lname = H.get('slim_attention_lname', 'Mixed_3b')
26 | early_feat = T[attention_lname]
27 |
28 | return coarse_feat, early_feat
29 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/hungarian/hungarian.cpp:
--------------------------------------------------------------------------------
1 | /********************************************************************
2 | ********************************************************************
3 | **
4 | ** libhungarian by Cyrill Stachniss, 2004
5 | **
6 | **
7 | ** Solving the Minimum Assignment Problem using the
8 | ** Hungarian Method.
9 | **
10 | ** ** This file may be freely copied and distributed! **
11 | **
12 | ** Parts of the used code was originally provided by the
13 | ** "Stanford GraphGase", but I made changes to this code.
14 | ** As asked by the copyright node of the "Stanford GraphGase",
15 | ** I hereby proclaim that this file are *NOT* part of the
16 | ** "Stanford GraphGase" distrubition!
17 | **
18 | ** This file is distributed in the hope that it will be useful,
19 | ** but WITHOUT ANY WARRANTY; without even the implied
20 | ** warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
21 | ** PURPOSE.
22 | **
23 | ********************************************************************
24 | ********************************************************************/
25 |
26 |
27 | #include
28 | #include
29 | #include "hungarian.hpp"
30 |
31 | #define INF (0x7FFFFFFF)
32 | #define verbose (0)
33 |
34 | #define hungarian_test_alloc(X) do {if ((void *)(X) == NULL) fprintf(stderr, "Out of memory in %s, (%s, line %d).\n", __FUNCTION__, __FILE__, __LINE__); } while (0)
35 |
36 | int** array_to_matrix(int* m, int rows, int cols) {
37 | int i,j;
38 | int** r;
39 | r = (int**)calloc(rows,sizeof(int*));
40 | for(i=0;iassignment, p->num_rows, p->num_cols) ;
64 | }
65 |
66 | void hungarian_print_costmatrix(hungarian_problem_t* p) {
67 | hungarian_print_matrix(p->cost, p->num_rows, p->num_cols) ;
68 | }
69 |
70 | void hungarian_print_status(hungarian_problem_t* p) {
71 |
72 | fprintf(stderr,"cost:\n");
73 | hungarian_print_costmatrix(p);
74 |
75 | fprintf(stderr,"assignment:\n");
76 | hungarian_print_assignment(p);
77 |
78 | }
79 |
80 | int hungarian_imax(int a, int b) {
81 | return (anum_rows = rows;
99 | p->num_cols = cols;
100 |
101 | p->cost = (int**)calloc(rows,sizeof(int*));
102 | hungarian_test_alloc(p->cost);
103 | p->assignment = (int**)calloc(rows,sizeof(int*));
104 | hungarian_test_alloc(p->assignment);
105 |
106 | for(i=0; inum_rows; i++) {
107 | p->cost[i] = (int*)calloc(cols,sizeof(int));
108 | hungarian_test_alloc(p->cost[i]);
109 | p->assignment[i] = (int*)calloc(cols,sizeof(int));
110 | hungarian_test_alloc(p->assignment[i]);
111 | for(j=0; jnum_cols; j++) {
112 | p->cost[i][j] = (i < org_rows && j < org_cols) ? cost_matrix[i][j] : 0;
113 | p->assignment[i][j] = 0;
114 |
115 | if (max_cost < p->cost[i][j])
116 | max_cost = p->cost[i][j];
117 | }
118 | }
119 |
120 |
121 | if (mode == HUNGARIAN_MODE_MAXIMIZE_UTIL) {
122 | for(i=0; inum_rows; i++) {
123 | for(j=0; jnum_cols; j++) {
124 | p->cost[i][j] = max_cost - p->cost[i][j];
125 | }
126 | }
127 | }
128 | else if (mode == HUNGARIAN_MODE_MINIMIZE_COST) {
129 | // nothing to do
130 | }
131 | else
132 | fprintf(stderr,"%s: unknown mode. Mode was set to HUNGARIAN_MODE_MINIMIZE_COST !\n", __FUNCTION__);
133 |
134 | return rows;
135 | }
136 |
137 |
138 |
139 |
140 | void hungarian_free(hungarian_problem_t* p) {
141 | int i;
142 | for(i=0; inum_rows; i++) {
143 | free(p->cost[i]);
144 | free(p->assignment[i]);
145 | }
146 | free(p->cost);
147 | free(p->assignment);
148 | p->cost = NULL;
149 | p->assignment = NULL;
150 | }
151 |
152 |
153 |
154 | void hungarian_solve(hungarian_problem_t* p)
155 | {
156 | int i, j, m, n, k, l, s, t, q, unmatched, cost;
157 | int* col_mate;
158 | int* row_mate;
159 | int* parent_row;
160 | int* unchosen_row;
161 | int* row_dec;
162 | int* col_inc;
163 | int* slack;
164 | int* slack_row;
165 |
166 | cost=0;
167 | m =p->num_rows;
168 | n =p->num_cols;
169 |
170 | col_mate = (int*)calloc(p->num_rows,sizeof(int));
171 | hungarian_test_alloc(col_mate);
172 | unchosen_row = (int*)calloc(p->num_rows,sizeof(int));
173 | hungarian_test_alloc(unchosen_row);
174 | row_dec = (int*)calloc(p->num_rows,sizeof(int));
175 | hungarian_test_alloc(row_dec);
176 | slack_row = (int*)calloc(p->num_rows,sizeof(int));
177 | hungarian_test_alloc(slack_row);
178 |
179 | row_mate = (int*)calloc(p->num_cols,sizeof(int));
180 | hungarian_test_alloc(row_mate);
181 | parent_row = (int*)calloc(p->num_cols,sizeof(int));
182 | hungarian_test_alloc(parent_row);
183 | col_inc = (int*)calloc(p->num_cols,sizeof(int));
184 | hungarian_test_alloc(col_inc);
185 | slack = (int*)calloc(p->num_cols,sizeof(int));
186 | hungarian_test_alloc(slack);
187 |
188 | for (i=0;inum_rows;i++) {
189 | col_mate[i]=0;
190 | unchosen_row[i]=0;
191 | row_dec[i]=0;
192 | slack_row[i]=0;
193 | }
194 | for (j=0;jnum_cols;j++) {
195 | row_mate[j]=0;
196 | parent_row[j] = 0;
197 | col_inc[j]=0;
198 | slack[j]=0;
199 | }
200 |
201 | for (i=0;inum_rows;++i)
202 | for (j=0;jnum_cols;++j)
203 | p->assignment[i][j]=HUNGARIAN_NOT_ASSIGNED;
204 |
205 | // Begin subtract column minima in order to start with lots of zeroes 12
206 | if (verbose)
207 | fprintf(stderr, "Using heuristic\n");
208 | for (l=0;lcost[0][l];
211 | for (k=1;kcost[k][l]cost[k][l];
214 | cost+=s;
215 | if (s!=0)
216 | for (k=0;kcost[k][l]-=s;
218 | }
219 | // End subtract column minima in order to start with lots of zeroes 12
220 |
221 | // Begin initial state 16
222 | t=0;
223 | for (l=0;lcost[k][0];
233 | for (l=1;lcost[k][l]cost[k][l];
236 | row_dec[k]=s;
237 | for (l=0;lcost[k][l] && row_mate[l]<0)
239 | {
240 | col_mate[k]=l;
241 | row_mate[l]=k;
242 | if (verbose)
243 | fprintf(stderr, "matching col %d==row %d\n",l,k);
244 | goto row_done;
245 | }
246 | col_mate[k]= -1;
247 | if (verbose)
248 | fprintf(stderr, "node %d: unmatched row %d\n",t,k);
249 | unchosen_row[t++]=k;
250 | row_done:
251 | ;
252 | }
253 | // End initial state 16
254 |
255 | // Begin Hungarian algorithm 18
256 | if (t==0)
257 | goto done;
258 | unmatched=t;
259 | while (1)
260 | {
261 | if (verbose)
262 | fprintf(stderr, "Matched %d rows.\n",m-t);
263 | q=0;
264 | while (1)
265 | {
266 | while (qcost[k][l]-s+col_inc[l];
277 | if (delcost[k][l]cost[k][l]!=row_dec[k]-col_inc[l])
388 | exit(0);
389 | }
390 | k=0;
391 | for (l=0;lm)
395 | exit(0);
396 | // End doublecheck the solution 23
397 | // End Hungarian algorithm 18
398 |
399 | for (i=0;iassignment[i][col_mate[i]]=HUNGARIAN_ASSIGNED;
402 | /*TRACE("%d - %d\n", i, col_mate[i]);*/
403 | }
404 | for (k=0;kcost[k][l]-row_dec[k]+col_inc[l]);*/
409 | p->cost[k][l]=p->cost[k][l]-row_dec[k]+col_inc[l];
410 | }
411 | /*TRACE("\n");*/
412 | }
413 | for (i=0;i (self.width + other.width) / 1.5:
11 | return False
12 | elif abs(self.cy - other.cy) > (self.height + other.height) / 2.0:
13 | return False
14 | else:
15 | return True
16 | def distance(self, other):
17 | return sum(map(abs, [self.cx - other.cx, self.cy - other.cy,
18 | self.width - other.width, self.height - other.height]))
19 | def intersection(self, other):
20 | left = max(self.cx - self.width/2., other.cx - other.width/2.)
21 | right = min(self.cx + self.width/2., other.cx + other.width/2.)
22 | width = max(right - left, 0)
23 | top = max(self.cy - self.height/2., other.cy - other.height/2.)
24 | bottom = min(self.cy + self.height/2., other.cy + other.height/2.)
25 | height = max(bottom - top, 0)
26 | return width * height
27 | def area(self):
28 | return self.height * self.width
29 | def union(self, other):
30 | return self.area() + other.area() - self.intersection(other)
31 | def iou(self, other):
32 | return self.intersection(other) / self.union(other)
33 | def __eq__(self, other):
34 | return (self.cx == other.cx and
35 | self.cy == other.cy and
36 | self.width == other.width and
37 | self.height == other.height and
38 | self.confidence == other.confidence)
39 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/slim_nets/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/bethesirius/ChosunTruck/889644385ce57f971ec2921f006fbb0a167e6f1e/linux/tensorbox/utils/slim_nets/__init__.py
--------------------------------------------------------------------------------
/linux/tensorbox/utils/slim_nets/resnet_utils.py:
--------------------------------------------------------------------------------
1 | # Copyright 2016 The TensorFlow Authors. All Rights Reserved.
2 | #
3 | # Licensed under the Apache License, Version 2.0 (the "License");
4 | # you may not use this file except in compliance with the License.
5 | # You may obtain a copy of the License at
6 | #
7 | # http://www.apache.org/licenses/LICENSE-2.0
8 | #
9 | # Unless required by applicable law or agreed to in writing, software
10 | # distributed under the License is distributed on an "AS IS" BASIS,
11 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 | # See the License for the specific language governing permissions and
13 | # limitations under the License.
14 | # ==============================================================================
15 | """Contains building blocks for various versions of Residual Networks.
16 |
17 | Residual networks (ResNets) were proposed in:
18 | Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
19 | Deep Residual Learning for Image Recognition. arXiv:1512.03385, 2015
20 |
21 | More variants were introduced in:
22 | Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
23 | Identity Mappings in Deep Residual Networks. arXiv: 1603.05027, 2016
24 |
25 | We can obtain different ResNet variants by changing the network depth, width,
26 | and form of residual unit. This module implements the infrastructure for
27 | building them. Concrete ResNet units and full ResNet networks are implemented in
28 | the accompanying resnet_v1.py and resnet_v2.py modules.
29 |
30 | Compared to https://github.com/KaimingHe/deep-residual-networks, in the current
31 | implementation we subsample the output activations in the last residual unit of
32 | each block, instead of subsampling the input activations in the first residual
33 | unit of each block. The two implementations give identical results but our
34 | implementation is more memory efficient.
35 | """
36 |
37 | from __future__ import absolute_import
38 | from __future__ import division
39 | from __future__ import print_function
40 |
41 | import collections
42 |
43 | from tensorflow.contrib import layers as layers_lib
44 | from tensorflow.contrib.framework.python.ops import add_arg_scope
45 | from tensorflow.contrib.framework.python.ops import arg_scope
46 | from tensorflow.contrib.layers.python.layers import initializers
47 | from tensorflow.contrib.layers.python.layers import layers
48 | from tensorflow.contrib.layers.python.layers import regularizers
49 | from tensorflow.contrib.layers.python.layers import utils
50 | from tensorflow.python.framework import ops
51 | from tensorflow.python.ops import array_ops
52 | from tensorflow.python.ops import nn_ops
53 | from tensorflow.python.ops import variable_scope
54 |
55 |
56 | class Block(collections.namedtuple('Block', ['scope', 'unit_fn', 'args'])):
57 | """A named tuple describing a ResNet block.
58 |
59 | Its parts are:
60 | scope: The scope of the `Block`.
61 | unit_fn: The ResNet unit function which takes as input a `Tensor` and
62 | returns another `Tensor` with the output of the ResNet unit.
63 | args: A list of length equal to the number of units in the `Block`. The list
64 | contains one (depth, depth_bottleneck, stride) tuple for each unit in the
65 | block to serve as argument to unit_fn.
66 | """
67 |
68 |
69 | def subsample(inputs, factor, scope=None):
70 | """Subsamples the input along the spatial dimensions.
71 |
72 | Args:
73 | inputs: A `Tensor` of size [batch, height_in, width_in, channels].
74 | factor: The subsampling factor.
75 | scope: Optional variable_scope.
76 |
77 | Returns:
78 | output: A `Tensor` of size [batch, height_out, width_out, channels] with the
79 | input, either intact (if factor == 1) or subsampled (if factor > 1).
80 | """
81 | if factor == 1:
82 | return inputs
83 | else:
84 | return layers.max_pool2d(inputs, [1, 1], stride=factor, scope=scope)
85 |
86 |
87 | def conv2d_same(inputs, num_outputs, kernel_size, stride, rate=1, scope=None):
88 | """Strided 2-D convolution with 'SAME' padding.
89 |
90 | When stride > 1, then we do explicit zero-padding, followed by conv2d with
91 | 'VALID' padding.
92 |
93 | Note that
94 |
95 | net = conv2d_same(inputs, num_outputs, 3, stride=stride)
96 |
97 | is equivalent to
98 |
99 | net = tf.contrib.layers.conv2d(inputs, num_outputs, 3, stride=1,
100 | padding='SAME')
101 | net = subsample(net, factor=stride)
102 |
103 | whereas
104 |
105 | net = tf.contrib.layers.conv2d(inputs, num_outputs, 3, stride=stride,
106 | padding='SAME')
107 |
108 | is different when the input's height or width is even, which is why we add the
109 | current function. For more details, see ResnetUtilsTest.testConv2DSameEven().
110 |
111 | Args:
112 | inputs: A 4-D tensor of size [batch, height_in, width_in, channels].
113 | num_outputs: An integer, the number of output filters.
114 | kernel_size: An int with the kernel_size of the filters.
115 | stride: An integer, the output stride.
116 | rate: An integer, rate for atrous convolution.
117 | scope: Scope.
118 |
119 | Returns:
120 | output: A 4-D tensor of size [batch, height_out, width_out, channels] with
121 | the convolution output.
122 | """
123 | if stride == 1:
124 | return layers_lib.conv2d(
125 | inputs,
126 | num_outputs,
127 | kernel_size,
128 | stride=1,
129 | rate=rate,
130 | padding='SAME',
131 | scope=scope)
132 | else:
133 | kernel_size_effective = kernel_size + (kernel_size - 1) * (rate - 1)
134 | pad_total = kernel_size_effective - 1
135 | pad_beg = pad_total // 2
136 | pad_end = pad_total - pad_beg
137 | inputs = array_ops.pad(
138 | inputs, [[0, 0], [pad_beg, pad_end], [pad_beg, pad_end], [0, 0]])
139 | return layers_lib.conv2d(
140 | inputs,
141 | num_outputs,
142 | kernel_size,
143 | stride=stride,
144 | rate=rate,
145 | padding='VALID',
146 | scope=scope)
147 |
148 |
149 | @add_arg_scope
150 | def stack_blocks_dense(net,
151 | blocks,
152 | output_stride=None,
153 | outputs_collections=None):
154 | """Stacks ResNet `Blocks` and controls output feature density.
155 |
156 | First, this function creates scopes for the ResNet in the form of
157 | 'block_name/unit_1', 'block_name/unit_2', etc.
158 |
159 | Second, this function allows the user to explicitly control the ResNet
160 | output_stride, which is the ratio of the input to output spatial resolution.
161 | This is useful for dense prediction tasks such as semantic segmentation or
162 | object detection.
163 |
164 | Most ResNets consist of 4 ResNet blocks and subsample the activations by a
165 | factor of 2 when transitioning between consecutive ResNet blocks. This results
166 | to a nominal ResNet output_stride equal to 8. If we set the output_stride to
167 | half the nominal network stride (e.g., output_stride=4), then we compute
168 | responses twice.
169 |
170 | Control of the output feature density is implemented by atrous convolution.
171 |
172 | Args:
173 | net: A `Tensor` of size [batch, height, width, channels].
174 | blocks: A list of length equal to the number of ResNet `Blocks`. Each
175 | element is a ResNet `Block` object describing the units in the `Block`.
176 | output_stride: If `None`, then the output will be computed at the nominal
177 | network stride. If output_stride is not `None`, it specifies the requested
178 | ratio of input to output spatial resolution, which needs to be equal to
179 | the product of unit strides from the start up to some level of the ResNet.
180 | For example, if the ResNet employs units with strides 1, 2, 1, 3, 4, 1,
181 | then valid values for the output_stride are 1, 2, 6, 24 or None (which
182 | is equivalent to output_stride=24).
183 | outputs_collections: Collection to add the ResNet block outputs.
184 |
185 | Returns:
186 | net: Output tensor with stride equal to the specified output_stride.
187 |
188 | Raises:
189 | ValueError: If the target output_stride is not valid.
190 | """
191 | # The current_stride variable keeps track of the effective stride of the
192 | # activations. This allows us to invoke atrous convolution whenever applying
193 | # the next residual unit would result in the activations having stride larger
194 | # than the target output_stride.
195 | current_stride = 1
196 |
197 | # The atrous convolution rate parameter.
198 | rate = 1
199 |
200 | for block in blocks:
201 | with variable_scope.variable_scope(block.scope, 'block', [net]) as sc:
202 | for i, unit in enumerate(block.args):
203 | if output_stride is not None and current_stride > output_stride:
204 | raise ValueError('The target output_stride cannot be reached.')
205 |
206 | with variable_scope.variable_scope('unit_%d' % (i + 1), values=[net]):
207 | unit_depth, unit_depth_bottleneck, unit_stride = unit
208 |
209 | # If we have reached the target output_stride, then we need to employ
210 | # atrous convolution with stride=1 and multiply the atrous rate by the
211 | # current unit's stride for use in subsequent layers.
212 | if output_stride is not None and current_stride == output_stride:
213 | net = block.unit_fn(
214 | net,
215 | depth=unit_depth,
216 | depth_bottleneck=unit_depth_bottleneck,
217 | stride=1,
218 | rate=rate)
219 | rate *= unit_stride
220 |
221 | else:
222 | net = block.unit_fn(
223 | net,
224 | depth=unit_depth,
225 | depth_bottleneck=unit_depth_bottleneck,
226 | stride=unit_stride,
227 | rate=1)
228 | current_stride *= unit_stride
229 | net = utils.collect_named_outputs(outputs_collections, sc.name, net)
230 |
231 | if output_stride is not None and current_stride != output_stride:
232 | raise ValueError('The target output_stride cannot be reached.')
233 |
234 | return net
235 |
236 |
237 | def resnet_arg_scope(is_training=True,
238 | weight_decay=0.0001,
239 | batch_norm_decay=0.997,
240 | batch_norm_epsilon=1e-5,
241 | batch_norm_scale=True):
242 | """Defines the default ResNet arg scope.
243 |
244 | TODO(gpapan): The batch-normalization related default values above are
245 | appropriate for use in conjunction with the reference ResNet models
246 | released at https://github.com/KaimingHe/deep-residual-networks. When
247 | training ResNets from scratch, they might need to be tuned.
248 |
249 | Args:
250 | is_training: Whether or not we are training the parameters in the batch
251 | normalization layers of the model.
252 | weight_decay: The weight decay to use for regularizing the model.
253 | batch_norm_decay: The moving average decay when estimating layer activation
254 | statistics in batch normalization.
255 | batch_norm_epsilon: Small constant to prevent division by zero when
256 | normalizing activations by their variance in batch normalization.
257 | batch_norm_scale: If True, uses an explicit `gamma` multiplier to scale the
258 | activations in the batch normalization layer.
259 |
260 | Returns:
261 | An `arg_scope` to use for the resnet models.
262 | """
263 | batch_norm_params = {
264 | 'is_training': is_training,
265 | 'decay': batch_norm_decay,
266 | 'epsilon': batch_norm_epsilon,
267 | 'scale': batch_norm_scale,
268 | 'updates_collections': ops.GraphKeys.UPDATE_OPS,
269 | }
270 |
271 | with arg_scope(
272 | [layers_lib.conv2d],
273 | weights_regularizer=regularizers.l2_regularizer(weight_decay),
274 | weights_initializer=initializers.variance_scaling_initializer(),
275 | activation_fn=nn_ops.relu,
276 | normalizer_fn=layers.batch_norm,
277 | normalizer_params=batch_norm_params):
278 | with arg_scope([layers.batch_norm], **batch_norm_params):
279 | # The following implies padding='SAME' for pool1, which makes feature
280 | # alignment easier for dense prediction tasks. This is also used in
281 | # https://github.com/facebook/fb.resnet.torch. However the accompanying
282 | # code of 'Deep Residual Learning for Image Recognition' uses
283 | # padding='VALID' for pool1. You can switch to that choice by setting
284 | # tf.contrib.framework.arg_scope([tf.contrib.layers.max_pool2d], padding='VALID').
285 | with arg_scope([layers.max_pool2d], padding='SAME') as arg_sc:
286 | return arg_sc
287 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/slim_nets/resnet_v1.py:
--------------------------------------------------------------------------------
1 | # Copyright 2016 The TensorFlow Authors. All Rights Reserved.
2 | #
3 | # Licensed under the Apache License, Version 2.0 (the "License");
4 | # you may not use this file except in compliance with the License.
5 | # You may obtain a copy of the License at
6 | #
7 | # http://www.apache.org/licenses/LICENSE-2.0
8 | #
9 | # Unless required by applicable law or agreed to in writing, software
10 | # distributed under the License is distributed on an "AS IS" BASIS,
11 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 | # See the License for the specific language governing permissions and
13 | # limitations under the License.
14 | # ==============================================================================
15 | """Contains definitions for the original form of Residual Networks.
16 |
17 | The 'v1' residual networks (ResNets) implemented in this module were proposed
18 | by:
19 | [1] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
20 | Deep Residual Learning for Image Recognition. arXiv:1512.03385
21 |
22 | Other variants were introduced in:
23 | [2] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
24 | Identity Mappings in Deep Residual Networks. arXiv: 1603.05027
25 |
26 | The networks defined in this module utilize the bottleneck building block of
27 | [1] with projection shortcuts only for increasing depths. They employ batch
28 | normalization *after* every weight layer. This is the architecture used by
29 | MSRA in the Imagenet and MSCOCO 2016 competition models ResNet-101 and
30 | ResNet-152. See [2; Fig. 1a] for a comparison between the current 'v1'
31 | architecture and the alternative 'v2' architecture of [2] which uses batch
32 | normalization *before* every weight layer in the so-called full pre-activation
33 | units.
34 |
35 | Typical use:
36 |
37 | from tensorflow.contrib.slim.nets import resnet_v1
38 |
39 | ResNet-101 for image classification into 1000 classes:
40 |
41 | # inputs has shape [batch, 224, 224, 3]
42 | with slim.arg_scope(resnet_v1.resnet_arg_scope()):
43 | net, end_points = resnet_v1.resnet_v1_101(inputs, 1000, is_training=False)
44 |
45 | ResNet-101 for semantic segmentation into 21 classes:
46 |
47 | # inputs has shape [batch, 513, 513, 3]
48 | with slim.arg_scope(resnet_v1.resnet_arg_scope()):
49 | net, end_points = resnet_v1.resnet_v1_101(inputs,
50 | 21,
51 | is_training=False,
52 | global_pool=False,
53 | output_stride=16)
54 | """
55 | from __future__ import absolute_import
56 | from __future__ import division
57 | from __future__ import print_function
58 |
59 | import tensorflow as tf
60 |
61 | from . import resnet_utils
62 |
63 |
64 | resnet_arg_scope = resnet_utils.resnet_arg_scope
65 | slim = tf.contrib.slim
66 |
67 |
68 | @slim.add_arg_scope
69 | def bottleneck(inputs, depth, depth_bottleneck, stride, rate=1,
70 | outputs_collections=None, scope=None):
71 | """Bottleneck residual unit variant with BN after convolutions.
72 |
73 | This is the original residual unit proposed in [1]. See Fig. 1(a) of [2] for
74 | its definition. Note that we use here the bottleneck variant which has an
75 | extra bottleneck layer.
76 |
77 | When putting together two consecutive ResNet blocks that use this unit, one
78 | should use stride = 2 in the last unit of the first block.
79 |
80 | Args:
81 | inputs: A tensor of size [batch, height, width, channels].
82 | depth: The depth of the ResNet unit output.
83 | depth_bottleneck: The depth of the bottleneck layers.
84 | stride: The ResNet unit's stride. Determines the amount of downsampling of
85 | the units output compared to its input.
86 | rate: An integer, rate for atrous convolution.
87 | outputs_collections: Collection to add the ResNet unit output.
88 | scope: Optional variable_scope.
89 |
90 | Returns:
91 | The ResNet unit's output.
92 | """
93 | with tf.variable_scope(scope, 'bottleneck_v1', [inputs]) as sc:
94 | depth_in = slim.utils.last_dimension(inputs.get_shape(), min_rank=4)
95 | if depth == depth_in:
96 | shortcut = resnet_utils.subsample(inputs, stride, 'shortcut')
97 | else:
98 | shortcut = slim.conv2d(inputs, depth, [1, 1], stride=stride,
99 | activation_fn=None, scope='shortcut')
100 |
101 | residual = slim.conv2d(inputs, depth_bottleneck, [1, 1], stride=1,
102 | scope='conv1')
103 | residual = resnet_utils.conv2d_same(residual, depth_bottleneck, 3, stride,
104 | rate=rate, scope='conv2')
105 | residual = slim.conv2d(residual, depth, [1, 1], stride=1,
106 | activation_fn=None, scope='conv3')
107 |
108 | output = tf.nn.relu(shortcut + residual)
109 |
110 | return slim.utils.collect_named_outputs(outputs_collections,
111 | sc.original_name_scope,
112 | output)
113 |
114 |
115 | def resnet_v1(inputs,
116 | blocks,
117 | num_classes=None,
118 | is_training=True,
119 | global_pool=True,
120 | output_stride=None,
121 | include_root_block=True,
122 | reuse=None,
123 | scope=None):
124 | """Generator for v1 ResNet models.
125 |
126 | This function generates a family of ResNet v1 models. See the resnet_v1_*()
127 | methods for specific model instantiations, obtained by selecting different
128 | block instantiations that produce ResNets of various depths.
129 |
130 | Training for image classification on Imagenet is usually done with [224, 224]
131 | inputs, resulting in [7, 7] feature maps at the output of the last ResNet
132 | block for the ResNets defined in [1] that have nominal stride equal to 32.
133 | However, for dense prediction tasks we advise that one uses inputs with
134 | spatial dimensions that are multiples of 32 plus 1, e.g., [321, 321]. In
135 | this case the feature maps at the ResNet output will have spatial shape
136 | [(height - 1) / output_stride + 1, (width - 1) / output_stride + 1]
137 | and corners exactly aligned with the input image corners, which greatly
138 | facilitates alignment of the features to the image. Using as input [225, 225]
139 | images results in [8, 8] feature maps at the output of the last ResNet block.
140 |
141 | For dense prediction tasks, the ResNet needs to run in fully-convolutional
142 | (FCN) mode and global_pool needs to be set to False. The ResNets in [1, 2] all
143 | have nominal stride equal to 32 and a good choice in FCN mode is to use
144 | output_stride=16 in order to increase the density of the computed features at
145 | small computational and memory overhead, cf. http://arxiv.org/abs/1606.00915.
146 |
147 | Args:
148 | inputs: A tensor of size [batch, height_in, width_in, channels].
149 | blocks: A list of length equal to the number of ResNet blocks. Each element
150 | is a resnet_utils.Block object describing the units in the block.
151 | num_classes: Number of predicted classes for classification tasks. If None
152 | we return the features before the logit layer.
153 | is_training: whether is training or not.
154 | global_pool: If True, we perform global average pooling before computing the
155 | logits. Set to True for image classification, False for dense prediction.
156 | output_stride: If None, then the output will be computed at the nominal
157 | network stride. If output_stride is not None, it specifies the requested
158 | ratio of input to output spatial resolution.
159 | include_root_block: If True, include the initial convolution followed by
160 | max-pooling, if False excludes it.
161 | reuse: whether or not the network and its variables should be reused. To be
162 | able to reuse 'scope' must be given.
163 | scope: Optional variable_scope.
164 |
165 | Returns:
166 | net: A rank-4 tensor of size [batch, height_out, width_out, channels_out].
167 | If global_pool is False, then height_out and width_out are reduced by a
168 | factor of output_stride compared to the respective height_in and width_in,
169 | else both height_out and width_out equal one. If num_classes is None, then
170 | net is the output of the last ResNet block, potentially after global
171 | average pooling. If num_classes is not None, net contains the pre-softmax
172 | activations.
173 | end_points: A dictionary from components of the network to the corresponding
174 | activation.
175 |
176 | Raises:
177 | ValueError: If the target output_stride is not valid.
178 | """
179 | with tf.variable_scope(scope, 'resnet_v1', [inputs], reuse=reuse) as sc:
180 | end_points_collection = sc.name + '_end_points'
181 | with slim.arg_scope([slim.conv2d, bottleneck,
182 | resnet_utils.stack_blocks_dense],
183 | outputs_collections=end_points_collection):
184 | with slim.arg_scope([slim.batch_norm], is_training=is_training):
185 | net = inputs
186 | if include_root_block:
187 | if output_stride is not None:
188 | if output_stride % 4 != 0:
189 | raise ValueError('The output_stride needs to be a multiple of 4.')
190 | output_stride /= 4
191 | net = resnet_utils.conv2d_same(net, 64, 7, stride=2, scope='conv1')
192 | net = slim.max_pool2d(net, [3, 3], stride=2, scope='pool1')
193 | net = resnet_utils.stack_blocks_dense(net, blocks, output_stride)
194 | if global_pool:
195 | # Global average pooling.
196 | net = tf.reduce_mean(net, [1, 2], name='pool5', keep_dims=True)
197 | if num_classes is not None:
198 | net = slim.conv2d(net, num_classes, [1, 1], activation_fn=None,
199 | normalizer_fn=None, scope='logits')
200 | # Convert end_points_collection into a dictionary of end_points.
201 | end_points = slim.utils.convert_collection_to_dict(end_points_collection)
202 | if num_classes is not None:
203 | end_points['predictions'] = slim.softmax(net, scope='predictions')
204 | return net, end_points
205 | resnet_v1.default_image_size = 224
206 |
207 |
208 | def resnet_v1_50(inputs,
209 | num_classes=None,
210 | is_training=True,
211 | global_pool=True,
212 | output_stride=None,
213 | reuse=None,
214 | scope='resnet_v1_50'):
215 | """ResNet-50 model of [1]. See resnet_v1() for arg and return description."""
216 | blocks = [
217 | resnet_utils.Block(
218 | 'block1', bottleneck, [(256, 64, 1)] * 2 + [(256, 64, 2)]),
219 | resnet_utils.Block(
220 | 'block2', bottleneck, [(512, 128, 1)] * 3 + [(512, 128, 2)]),
221 | resnet_utils.Block(
222 | 'block3', bottleneck, [(1024, 256, 1)] * 5 + [(1024, 256, 2)]),
223 | resnet_utils.Block(
224 | 'block4', bottleneck, [(2048, 512, 1)] * 3)
225 | ]
226 | return resnet_v1(inputs, blocks, num_classes, is_training,
227 | global_pool=global_pool, output_stride=output_stride,
228 | include_root_block=True, reuse=reuse, scope=scope)
229 |
230 |
231 | def resnet_v1_101(inputs,
232 | num_classes=None,
233 | is_training=True,
234 | global_pool=True,
235 | output_stride=None,
236 | reuse=None,
237 | scope='resnet_v1_101'):
238 | """ResNet-101 model of [1]. See resnet_v1() for arg and return description."""
239 | blocks = [
240 | resnet_utils.Block(
241 | 'block1', bottleneck, [(256, 64, 1)] * 2 + [(256, 64, 2)]),
242 | resnet_utils.Block(
243 | 'block2', bottleneck, [(512, 128, 1)] * 3 + [(512, 128, 2)]),
244 | resnet_utils.Block(
245 | 'block3', bottleneck, [(1024, 256, 1)] * 22 + [(1024, 256, 2)]),
246 | resnet_utils.Block(
247 | 'block4', bottleneck, [(2048, 512, 1)] * 3)
248 | ]
249 | return resnet_v1(inputs, blocks, num_classes, is_training,
250 | global_pool=global_pool, output_stride=output_stride,
251 | include_root_block=True, reuse=reuse, scope=scope)
252 |
253 |
254 | def resnet_v1_152(inputs,
255 | num_classes=None,
256 | is_training=True,
257 | global_pool=True,
258 | output_stride=None,
259 | reuse=None,
260 | scope='resnet_v1_152'):
261 | """ResNet-152 model of [1]. See resnet_v1() for arg and return description."""
262 | blocks = [
263 | resnet_utils.Block(
264 | 'block1', bottleneck, [(256, 64, 1)] * 2 + [(256, 64, 2)]),
265 | resnet_utils.Block(
266 | 'block2', bottleneck, [(512, 128, 1)] * 7 + [(512, 128, 2)]),
267 | resnet_utils.Block(
268 | 'block3', bottleneck, [(1024, 256, 1)] * 35 + [(1024, 256, 2)]),
269 | resnet_utils.Block(
270 | 'block4', bottleneck, [(2048, 512, 1)] * 3)]
271 | return resnet_v1(inputs, blocks, num_classes, is_training,
272 | global_pool=global_pool, output_stride=output_stride,
273 | include_root_block=True, reuse=reuse, scope=scope)
274 |
275 |
276 | def resnet_v1_200(inputs,
277 | num_classes=None,
278 | is_training=True,
279 | global_pool=True,
280 | output_stride=None,
281 | reuse=None,
282 | scope='resnet_v1_200'):
283 | """ResNet-200 model of [2]. See resnet_v1() for arg and return description."""
284 | blocks = [
285 | resnet_utils.Block(
286 | 'block1', bottleneck, [(256, 64, 1)] * 2 + [(256, 64, 2)]),
287 | resnet_utils.Block(
288 | 'block2', bottleneck, [(512, 128, 1)] * 23 + [(512, 128, 2)]),
289 | resnet_utils.Block(
290 | 'block3', bottleneck, [(1024, 256, 1)] * 35 + [(1024, 256, 2)]),
291 | resnet_utils.Block(
292 | 'block4', bottleneck, [(2048, 512, 1)] * 3)]
293 | return resnet_v1(inputs, blocks, num_classes, is_training,
294 | global_pool=global_pool, output_stride=output_stride,
295 | include_root_block=True, reuse=reuse, scope=scope)
296 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/stitch_rects.cpp:
--------------------------------------------------------------------------------
1 | #include
2 | #include
3 | #include
4 | #include
5 | #include
6 | #include
7 |
8 | #include "./hungarian/hungarian.hpp"
9 | #include "./stitch_rects.hpp"
10 |
11 | using std::vector;
12 |
13 | void filter_rects(const vector > >& all_rects,
14 | vector* stitched_rects,
15 | float threshold,
16 | float max_threshold,
17 | float tau,
18 | float conf_alpha) {
19 | const vector& accepted_rects = *stitched_rects;
20 | for (int i = 0; i < (int)all_rects.size(); ++i) {
21 | for (int j = 0; j < (int)all_rects[0].size(); ++j) {
22 | vector current_rects;
23 | for (int k = 0; k < (int)all_rects[i][j].size(); ++k) {
24 | if (all_rects[i][j][k].confidence_ * conf_alpha > threshold) {
25 | Rect r = Rect(all_rects[i][j][k]);
26 | r.confidence_ *= conf_alpha;
27 | r.true_confidence_ *= conf_alpha;
28 | current_rects.push_back(r);
29 | }
30 | }
31 |
32 | vector relevant_rects;
33 | for (int k = 0; k < (int)accepted_rects.size(); ++k) {
34 | for (int l = 0; l < (int)current_rects.size(); ++l) {
35 | if (accepted_rects[k].overlaps(current_rects[l], tau)) {
36 | relevant_rects.push_back(Rect(accepted_rects[k]));
37 | break;
38 | }
39 | }
40 | }
41 |
42 | if (relevant_rects.size() == 0 || current_rects.size() == 0) {
43 | for (int k = 0; k < (int)current_rects.size(); ++k) {
44 | stitched_rects->push_back(Rect(current_rects[k]));
45 | }
46 | continue;
47 | }
48 |
49 | int num_pred = MAX(current_rects.size(), relevant_rects.size());
50 |
51 | int int_cost[num_pred * num_pred];
52 | for (int k = 0; k < num_pred * num_pred; ++k) { int_cost[k] = 0; }
53 | for (int k = 0; k < (int)current_rects.size(); ++k) {
54 | for (int l = 0; l < (int)relevant_rects.size(); ++l) {
55 | int idx = k * num_pred + l;
56 | int cost = 10000;
57 | if (current_rects[k].overlaps(relevant_rects[l], tau)) {
58 | cost -= 1000;
59 | }
60 | cost += (int)(current_rects[k].distance(relevant_rects[l]) / 10.);
61 | int_cost[idx] = cost;
62 | }
63 | }
64 |
65 | std::vector assignment;
66 |
67 | hungarian_problem_t p;
68 | int** m = array_to_matrix(int_cost, num_pred, num_pred);
69 | hungarian_init(&p, m, num_pred, num_pred, HUNGARIAN_MODE_MINIMIZE_COST);
70 | hungarian_solve(&p);
71 | for (int i = 0; i < num_pred; ++i) {
72 | for (int j = 0; j < num_pred; ++j) {
73 | if (p.assignment[i][j] == HUNGARIAN_ASSIGNED) {
74 | assignment.push_back(j);
75 | }
76 | }
77 | }
78 | assert((int)assignment.size() == num_pred);
79 | hungarian_free(&p);
80 |
81 | for (int i = 0; i < num_pred; ++i) {
82 | free(m[i]);
83 | }
84 | free(m);
85 |
86 | vector bad;
87 | for (int k = 0; k < (int)assignment.size(); ++k) {
88 | if (k < (int)current_rects.size() && assignment[k] < (int)relevant_rects.size()) {
89 | Rect& c = current_rects[k];
90 | Rect& a = relevant_rects[assignment[k]];
91 | if (c.confidence_ > max_threshold) {
92 | bad.push_back(k);
93 | continue;
94 | }
95 | if (c.overlaps(a, tau)) {
96 | if (c.confidence_ > a.confidence_ && c.iou(a) > 0.7) {
97 | c.true_confidence_ = a.confidence_;
98 | stitched_rects->erase(std::find(stitched_rects->begin(), stitched_rects->end(), a));
99 | } else {
100 | bad.push_back(k);
101 | }
102 | }
103 | }
104 | }
105 |
106 | for (int k = 0; k < (int)current_rects.size(); ++k) {
107 | bool bad_contains_k = false;
108 | for (int l = 0; l < (int)bad.size(); ++l) {
109 | if (k == bad[l]) {
110 | bad_contains_k = true;
111 | break;
112 | }
113 | }
114 | if (!bad_contains_k) {
115 | stitched_rects->push_back(Rect(current_rects[k]));
116 | }
117 | }
118 | }
119 | }
120 | }
121 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/stitch_rects.hpp:
--------------------------------------------------------------------------------
1 | #ifndef STITCH_RECTS_HPP
2 | #define STITCH_RECTS_HPP
3 |
4 | #include
5 | #include
6 | #include
7 |
8 | #include "./hungarian/hungarian.hpp"
9 |
10 | #define MIN(a,b) (((a)<(b))?(a):(b))
11 | #define MAX(a,b) (((a)>(b))?(a):(b))
12 |
13 | using std::vector;
14 |
15 | class Rect {
16 | public:
17 | int cx_;
18 | int cy_;
19 | int width_;
20 | int height_;
21 | float confidence_;
22 | float true_confidence_;
23 |
24 | explicit Rect(int cx, int cy, int width, int height, float confidence) {
25 | cx_ = cx;
26 | cy_ = cy;
27 | width_ = width;
28 | height_ = height;
29 | confidence_ = confidence;
30 | true_confidence_ = confidence;
31 | }
32 |
33 | Rect(const Rect& other) {
34 | cx_ = other.cx_;
35 | cy_ = other.cy_;
36 | width_ = other.width_;
37 | height_ = other.height_;
38 | confidence_ = other.confidence_;
39 | true_confidence_ = other.true_confidence_;
40 | }
41 |
42 | bool overlaps(const Rect& other, float tau) const {
43 | if (fabs(cx_ - other.cx_) > (width_ + other.width_) / 1.5) {
44 | return false;
45 | } else if (fabs(cy_ - other.cy_) > (height_ + other.height_) / 2.0) {
46 | return false;
47 | } else {
48 | return iou(other) > tau;
49 | }
50 | }
51 |
52 | int distance(const Rect& other) const {
53 | return (fabs(cx_ - other.cx_) + fabs(cy_ - other.cy_) +
54 | fabs(width_ - other.width_) + fabs(height_ - other.height_));
55 | }
56 |
57 | float intersection(const Rect& other) const {
58 | int left = MAX(cx_ - width_ / 2., other.cx_ - other.width_ / 2.);
59 | int right = MIN(cx_ + width_ / 2., other.cx_ + other.width_ / 2.);
60 | int width = MAX(right - left, 0);
61 |
62 | int top = MAX(cy_ - height_ / 2., other.cy_ - other.height_ / 2.);
63 | int bottom = MIN(cy_ + height_ / 2., other.cy_ + other.height_ / 2.);
64 | int height = MAX(bottom - top, 0);
65 | return width * height;
66 | }
67 |
68 | int area() const {
69 | return height_ * width_;
70 | }
71 |
72 | float union_area(const Rect& other) const {
73 | return this->area() + other.area() - this->intersection(other);
74 | }
75 |
76 | float iou(const Rect& other) const {
77 | return this->intersection(other) / this->union_area(other);
78 | }
79 |
80 | bool operator==(const Rect& other) const {
81 | return (cx_ == other.cx_ &&
82 | cy_ == other.cy_ &&
83 | width_ == other.width_ &&
84 | height_ == other.height_ &&
85 | confidence_ == other.confidence_);
86 | }
87 | };
88 |
89 | void filter_rects(const vector > >& all_rects,
90 | vector* stitched_rects,
91 | float threshold,
92 | float max_threshold,
93 | float tau,
94 | float conf_alpha);
95 |
96 | #endif // STITCH_RECTS_HPP
97 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/stitch_wrapper.py:
--------------------------------------------------------------------------------
1 | print('ERROR: stitch_wrapper not yet compiled. Please run `cd /path/to/tensorbox/utils && make`')
2 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/stitch_wrapper.pyx:
--------------------------------------------------------------------------------
1 | from libcpp.vector cimport vector
2 | from libcpp.set cimport set
3 | from rect import Rect as PyRect
4 | cdef extern from "stitch_rects.hpp":
5 | cdef cppclass Rect:
6 | Rect(int cx, int cy, int width, int height, float confidence)
7 | int cx_
8 | int cy_
9 | int width_
10 | int height_
11 | float confidence_
12 | float true_confidence_
13 |
14 | cdef void filter_rects(vector[vector[vector[Rect] ] ]& all_rects,
15 | vector[Rect]* stitched_rects,
16 | float threshold,
17 | float max_threshold,
18 | float tau,
19 | float conf_alpha);
20 |
21 | def stitch_rects(all_rects, tau=0.25):
22 | """
23 | Implements the stitching procedure discussed in the paper.
24 | Complicated, but we find that it does better than simpler versions
25 | and generalizes well across widely varying box sizes.
26 |
27 | Input:
28 | all_rects : 2d grid with each cell containing a vector of PyRects
29 | """
30 | for row in all_rects:
31 | assert len(row) == len(all_rects[0])
32 |
33 | cdef vector[vector[vector[Rect]]] c_rects
34 | cdef vector[vector[Rect]] c_row
35 | cdef vector[Rect] c_column
36 | for i, row in enumerate(all_rects):
37 | c_rects.push_back(c_row)
38 | for j, column in enumerate(row):
39 | c_rects[i].push_back(c_column)
40 | for py_rect in column:
41 | c_rects[i][j].push_back(
42 | Rect(
43 | py_rect.cx,
44 | py_rect.cy,
45 | py_rect.width,
46 | py_rect.height,
47 | py_rect.confidence)
48 | )
49 |
50 | cdef vector[Rect] acc_rects;
51 |
52 | thresholds = [(.80, 1.0),
53 | (.70, 0.9),
54 | (.60, 0.8),
55 | (.50, 0.7),
56 | (.40, 0.6),
57 | (.30, 0.5),
58 | (.20, 0.4),
59 | (.10, 0.3),
60 | (.05, 0.2),
61 | (.02, 0.1),
62 | (.005, 0.04),
63 | (.001, 0.01),
64 | ]
65 | t_conf_alphas = [(tau, 1.0),
66 | #(1 - (1 - tau) * 0.75, 0.5),
67 | #(1 - (1 - tau) * 0.5, 0.1),
68 | #(1 - (1 - tau) * 0.25, 0.005),
69 | ]
70 | for t, conf_alpha in t_conf_alphas:
71 | for lower_t, upper_t in thresholds:
72 | if lower_t * conf_alpha > 0.0001:
73 | filter_rects(c_rects, &acc_rects, lower_t * conf_alpha,
74 | upper_t * conf_alpha, t, conf_alpha)
75 |
76 | py_acc_rects = []
77 | for i in range(acc_rects.size()):
78 | acc_rect = PyRect(
79 | acc_rects[i].cx_,
80 | acc_rects[i].cy_,
81 | acc_rects[i].width_,
82 | acc_rects[i].height_,
83 | acc_rects[i].confidence_)
84 | acc_rect.true_confidence = acc_rects[i].true_confidence_
85 | py_acc_rects.append(acc_rect)
86 | return py_acc_rects
87 |
--------------------------------------------------------------------------------
/linux/tensorbox/utils/train_utils.py:
--------------------------------------------------------------------------------
1 | import numpy as np
2 | import random
3 | import json
4 | import os
5 | import cv2
6 | import itertools
7 | from scipy.misc import imread, imresize
8 | import tensorflow as tf
9 |
10 | from data_utils import (annotation_jitter, annotation_to_h5)
11 | from utils.annolist import AnnotationLib as al
12 | from rect import Rect
13 | from utils import tf_concat
14 |
15 | def rescale_boxes(current_shape, anno, target_height, target_width):
16 | x_scale = target_width / float(current_shape[1])
17 | y_scale = target_height / float(current_shape[0])
18 | for r in anno.rects:
19 | assert r.x1 < r.x2
20 | r.x1 *= x_scale
21 | r.x2 *= x_scale
22 | assert r.y1 < r.y2
23 | r.y1 *= y_scale
24 | r.y2 *= y_scale
25 | return anno
26 |
27 | def load_idl_tf(idlfile, H, jitter):
28 | """Take the idlfile and net configuration and create a generator
29 | that outputs a jittered version of a random image from the annolist
30 | that is mean corrected."""
31 |
32 | annolist = al.parse(idlfile)
33 | annos = []
34 | for anno in annolist:
35 | anno.imageName = os.path.join(
36 | os.path.dirname(os.path.realpath(idlfile)), anno.imageName)
37 | annos.append(anno)
38 | random.seed(0)
39 | if H['data']['truncate_data']:
40 | annos = annos[:10]
41 | for epoch in itertools.count():
42 | random.shuffle(annos)
43 | for anno in annos:
44 | I = imread(anno.imageName)
45 | #Skip Greyscale images
46 | if len(I.shape) < 3:
47 | continue
48 | if I.shape[2] == 4:
49 | I = I[:, :, :3]
50 | if I.shape[0] != H["image_height"] or I.shape[1] != H["image_width"]:
51 | if epoch == 0:
52 | anno = rescale_boxes(I.shape, anno, H["image_height"], H["image_width"])
53 | I = imresize(I, (H["image_height"], H["image_width"]), interp='cubic')
54 | if jitter:
55 | jitter_scale_min=0.9
56 | jitter_scale_max=1.1
57 | jitter_offset=16
58 | I, anno = annotation_jitter(I,
59 | anno, target_width=H["image_width"],
60 | target_height=H["image_height"],
61 | jitter_scale_min=jitter_scale_min,
62 | jitter_scale_max=jitter_scale_max,
63 | jitter_offset=jitter_offset)
64 |
65 | boxes, flags = annotation_to_h5(H,
66 | anno,
67 | H["grid_width"],
68 | H["grid_height"],
69 | H["rnn_len"])
70 |
71 | yield {"image": I, "boxes": boxes, "flags": flags}
72 |
73 | def make_sparse(n, d):
74 | v = np.zeros((d,), dtype=np.float32)
75 | v[n] = 1.
76 | return v
77 |
78 | def load_data_gen(H, phase, jitter):
79 | grid_size = H['grid_width'] * H['grid_height']
80 |
81 | data = load_idl_tf(H["data"]['%s_idl' % phase], H, jitter={'train': jitter, 'test': False}[phase])
82 |
83 | for d in data:
84 | output = {}
85 |
86 | rnn_len = H["rnn_len"]
87 | flags = d['flags'][0, :, 0, 0:rnn_len, 0]
88 | boxes = np.transpose(d['boxes'][0, :, :, 0:rnn_len, 0], (0, 2, 1))
89 | assert(flags.shape == (grid_size, rnn_len))
90 | assert(boxes.shape == (grid_size, rnn_len, 4))
91 |
92 | output['image'] = d['image']
93 | output['confs'] = np.array([[make_sparse(int(detection), d=H['num_classes']) for detection in cell] for cell in flags])
94 | output['boxes'] = boxes
95 | output['flags'] = flags
96 |
97 | yield output
98 |
99 | def add_rectangles(H, orig_image, confidences, boxes, use_stitching=False, rnn_len=1, min_conf=0.1, show_removed=True, tau=0.25, show_suppressed=True):
100 | image = np.copy(orig_image[0])
101 | num_cells = H["grid_height"] * H["grid_width"]
102 | boxes_r = np.reshape(boxes, (-1,
103 | H["grid_height"],
104 | H["grid_width"],
105 | rnn_len,
106 | 4))
107 | confidences_r = np.reshape(confidences, (-1,
108 | H["grid_height"],
109 | H["grid_width"],
110 | rnn_len,
111 | H['num_classes']))
112 | cell_pix_size = H['region_size']
113 | all_rects = [[[] for _ in range(H["grid_width"])] for _ in range(H["grid_height"])]
114 | for n in range(rnn_len):
115 | for y in range(H["grid_height"]):
116 | for x in range(H["grid_width"]):
117 | bbox = boxes_r[0, y, x, n, :]
118 | abs_cx = int(bbox[0]) + cell_pix_size/2 + cell_pix_size * x
119 | abs_cy = int(bbox[1]) + cell_pix_size/2 + cell_pix_size * y
120 | w = bbox[2]
121 | h = bbox[3]
122 | conf = np.max(confidences_r[0, y, x, n, 1:])
123 | all_rects[y][x].append(Rect(abs_cx,abs_cy,w,h,conf))
124 |
125 | all_rects_r = [r for row in all_rects for cell in row for r in cell]
126 | if use_stitching:
127 | from stitch_wrapper import stitch_rects
128 | acc_rects = stitch_rects(all_rects, tau)
129 | else:
130 | acc_rects = all_rects_r
131 |
132 |
133 | if show_suppressed:
134 | pairs = [(all_rects_r, (255, 0, 0))]
135 | else:
136 | pairs = []
137 | pairs.append((acc_rects, (0, 255, 0)))
138 | for rect_set, color in pairs:
139 | for rect in rect_set:
140 | if rect.confidence > min_conf:
141 | cv2.rectangle(image,
142 | (rect.cx-int(rect.width/2), rect.cy-int(rect.height/2)),
143 | (rect.cx+int(rect.width/2), rect.cy+int(rect.height/2)),
144 | color,
145 | 2)
146 |
147 | rects = []
148 | for rect in acc_rects:
149 | r = al.AnnoRect()
150 | r.x1 = rect.cx - rect.width/2.
151 | r.x2 = rect.cx + rect.width/2.
152 | r.y1 = rect.cy - rect.height/2.
153 | r.y2 = rect.cy + rect.height/2.
154 | r.score = rect.true_confidence
155 | rects.append(r)
156 |
157 | return image, rects
158 |
159 | def to_x1y1x2y2(box):
160 | w = tf.maximum(box[:, 2:3], 1)
161 | h = tf.maximum(box[:, 3:4], 1)
162 | x1 = box[:, 0:1] - w / 2
163 | x2 = box[:, 0:1] + w / 2
164 | y1 = box[:, 1:2] - h / 2
165 | y2 = box[:, 1:2] + h / 2
166 | return tf_concat(1, [x1, y1, x2, y2])
167 |
168 | def intersection(box1, box2):
169 | x1_max = tf.maximum(box1[:, 0], box2[:, 0])
170 | y1_max = tf.maximum(box1[:, 1], box2[:, 1])
171 | x2_min = tf.minimum(box1[:, 2], box2[:, 2])
172 | y2_min = tf.minimum(box1[:, 3], box2[:, 3])
173 |
174 | x_diff = tf.maximum(x2_min - x1_max, 0)
175 | y_diff = tf.maximum(y2_min - y1_max, 0)
176 |
177 | return x_diff * y_diff
178 |
179 | def area(box):
180 | x_diff = tf.maximum(box[:, 2] - box[:, 0], 0)
181 | y_diff = tf.maximum(box[:, 3] - box[:, 1], 0)
182 | return x_diff * y_diff
183 |
184 | def union(box1, box2):
185 | return area(box1) + area(box2) - intersection(box1, box2)
186 |
187 | def iou(box1, box2):
188 | return intersection(box1, box2) / union(box1, box2)
189 |
190 | def to_idx(vec, w_shape):
191 | '''
192 | vec = (idn, idh, idw)
193 | w_shape = [n, h, w, c]
194 | '''
195 | return vec[:, 2] + w_shape[2] * (vec[:, 1] + w_shape[1] * vec[:, 0])
196 |
197 | def interp(w, i, channel_dim):
198 | '''
199 | Input:
200 | w: A 4D block tensor of shape (n, h, w, c)
201 | i: A list of 3-tuples [(x_1, y_1, z_1), (x_2, y_2, z_2), ...],
202 | each having type (int, float, float)
203 |
204 | The 4D block represents a batch of 3D image feature volumes with c channels.
205 | The input i is a list of points to index into w via interpolation. Direct
206 | indexing is not possible due to y_1 and z_1 being float values.
207 | Output:
208 | A list of the values: [
209 | w[x_1, y_1, z_1, :]
210 | w[x_2, y_2, z_2, :]
211 | ...
212 | w[x_k, y_k, z_k, :]
213 | ]
214 | of the same length == len(i)
215 | '''
216 | w_as_vector = tf.reshape(w, [-1, channel_dim]) # gather expects w to be 1-d
217 | upper_l = tf.to_int32(tf_concat(1, [i[:, 0:1], tf.floor(i[:, 1:2]), tf.floor(i[:, 2:3])]))
218 | upper_r = tf.to_int32(tf_concat(1, [i[:, 0:1], tf.floor(i[:, 1:2]), tf.ceil(i[:, 2:3])]))
219 | lower_l = tf.to_int32(tf_concat(1, [i[:, 0:1], tf.ceil(i[:, 1:2]), tf.floor(i[:, 2:3])]))
220 | lower_r = tf.to_int32(tf_concat(1, [i[:, 0:1], tf.ceil(i[:, 1:2]), tf.ceil(i[:, 2:3])]))
221 |
222 | upper_l_idx = to_idx(upper_l, tf.shape(w))
223 | upper_r_idx = to_idx(upper_r, tf.shape(w))
224 | lower_l_idx = to_idx(lower_l, tf.shape(w))
225 | lower_r_idx = to_idx(lower_r, tf.shape(w))
226 |
227 | upper_l_value = tf.gather(w_as_vector, upper_l_idx)
228 | upper_r_value = tf.gather(w_as_vector, upper_r_idx)
229 | lower_l_value = tf.gather(w_as_vector, lower_l_idx)
230 | lower_r_value = tf.gather(w_as_vector, lower_r_idx)
231 |
232 | alpha_lr = tf.expand_dims(i[:, 2] - tf.floor(i[:, 2]), 1)
233 | alpha_ud = tf.expand_dims(i[:, 1] - tf.floor(i[:, 1]), 1)
234 |
235 | upper_value = (1 - alpha_lr) * upper_l_value + (alpha_lr) * upper_r_value
236 | lower_value = (1 - alpha_lr) * lower_l_value + (alpha_lr) * lower_r_value
237 | value = (1 - alpha_ud) * upper_value + (alpha_ud) * lower_value
238 | return value
239 |
240 | def bilinear_select(H, pred_boxes, early_feat, early_feat_channels, w_offset, h_offset):
241 | '''
242 | Function used for rezooming high level feature maps. Uses bilinear interpolation
243 | to select all channels at index (x, y) for a high level feature map, where x and y are floats.
244 | '''
245 | grid_size = H['grid_width'] * H['grid_height']
246 | outer_size = grid_size * H['batch_size']
247 |
248 | fine_stride = 8. # pixels per 60x80 grid cell in 480x640 image
249 | coarse_stride = H['region_size'] # pixels per 15x20 grid cell in 480x640 image
250 | batch_ids = []
251 | x_offsets = []
252 | y_offsets = []
253 | for n in range(H['batch_size']):
254 | for i in range(H['grid_height']):
255 | for j in range(H['grid_width']):
256 | for k in range(H['rnn_len']):
257 | batch_ids.append([n])
258 | x_offsets.append([coarse_stride / 2. + coarse_stride * j])
259 | y_offsets.append([coarse_stride / 2. + coarse_stride * i])
260 |
261 | batch_ids = tf.constant(batch_ids)
262 | x_offsets = tf.constant(x_offsets)
263 | y_offsets = tf.constant(y_offsets)
264 |
265 | pred_boxes_r = tf.reshape(pred_boxes, [outer_size * H['rnn_len'], 4])
266 | scale_factor = coarse_stride / fine_stride # scale difference between 15x20 and 60x80 features
267 |
268 | pred_x_center = (pred_boxes_r[:, 0:1] + w_offset * pred_boxes_r[:, 2:3] + x_offsets) / fine_stride
269 | pred_x_center_clip = tf.clip_by_value(pred_x_center,
270 | 0,
271 | scale_factor * H['grid_width'] - 1)
272 | pred_y_center = (pred_boxes_r[:, 1:2] + h_offset * pred_boxes_r[:, 3:4] + y_offsets) / fine_stride
273 | pred_y_center_clip = tf.clip_by_value(pred_y_center,
274 | 0,
275 | scale_factor * H['grid_height'] - 1)
276 |
277 | interp_indices = tf_concat(1, [tf.to_float(batch_ids), pred_y_center_clip, pred_x_center_clip])
278 | return interp_indices
279 |
--------------------------------------------------------------------------------
/windows/src/IPM.cpp:
--------------------------------------------------------------------------------
1 | #include "IPM.h"
2 |
3 | using namespace cv;
4 | using namespace std;
5 |
6 | // Public
7 | IPM::IPM( const cv::Size& _origSize, const cv::Size& _dstSize, const std::vector& _origPoints, const std::vector& _dstPoints )
8 | : m_origSize(_origSize), m_dstSize(_dstSize), m_origPoints(_origPoints), m_dstPoints(_dstPoints)
9 | {
10 | assert( m_origPoints.size() == 4 && m_dstPoints.size() == 4 && "Orig. points and Dst. points must vectors of 4 points" );
11 | m_H = getPerspectiveTransform( m_origPoints, m_dstPoints );
12 | m_H_inv = m_H.inv();
13 |
14 | createMaps();
15 | }
16 | void IPM::drawPoints( const std::vector& _points, cv::Mat& _img ) const
17 | {
18 | assert(_points.size() == 4);
19 |
20 | line(_img, Point(static_cast(_points[0].x), static_cast(_points[0].y)), Point(static_cast(_points[3].x), static_cast(_points[3].y)), CV_RGB( 205,205,0), 2);
21 | line(_img, Point(static_cast(_points[2].x), static_cast(_points[2].y)), Point(static_cast(_points[3].x), static_cast(_points[3].y)), CV_RGB( 205,205,0), 2);
22 | line(_img, Point(static_cast(_points[0].x), static_cast(_points[0].y)), Point(static_cast(_points[1].x), static_cast(_points[1].y)), CV_RGB( 205,205,0), 2);
23 | line(_img, Point(static_cast(_points[2].x), static_cast(_points[2].y)), Point(static_cast(_points[1].x), static_cast(_points[1].y)), CV_RGB( 205,205,0), 2);
24 | for(std::size_t i=0; i<_points.size(); i++)
25 | {
26 | circle(_img, Point(static_cast(_points[i].x), static_cast(_points[i].y)), 2, CV_RGB(238,238,0), -1);
27 | circle(_img, Point(static_cast(_points[i].x), static_cast(_points[i].y)), 5, CV_RGB(255,255,255), 2);
28 | }
29 | }
30 | void IPM::getPoints(vector& _origPts, vector& _ipmPts)
31 | {
32 | _origPts = m_origPoints;
33 | _ipmPts = m_dstPoints;
34 | }
35 | void IPM::applyHomography(const Mat& _inputImg, Mat& _dstImg, int _borderMode)
36 | {
37 | // Generate IPM image from src
38 | remap(_inputImg, _dstImg, m_mapX, m_mapY, INTER_LINEAR, _borderMode);//, BORDER_CONSTANT, Scalar(0,0,0,0));
39 | }
40 | void IPM::applyHomographyInv(const Mat& _inputImg, Mat& _dstImg, int _borderMode)
41 | {
42 | // Generate IPM image from src
43 | remap(_inputImg, _dstImg, m_mapX, m_mapY, INTER_LINEAR, _borderMode);//, BORDER_CONSTANT, Scalar(0,0,0,0));
44 | }
45 | Point2d IPM::applyHomography( const Point2d& _point )
46 | {
47 | return applyHomography( _point, m_H );
48 | }
49 | Point2d IPM::applyHomographyInv( const Point2d& _point )
50 | {
51 | return applyHomography( _point, m_H_inv );
52 | }
53 | Point2d IPM::applyHomography( const Point2d& _point, const Mat& _H )
54 | {
55 | Point2d ret = Point2d( -1, -1 );
56 |
57 | const double u = _H.at(0,0) * _point.x + _H.at(0,1) * _point.y + _H.at(0,2);
58 | const double v = _H.at(1,0) * _point.x + _H.at(1,1) * _point.y + _H.at(1,2);
59 | const double s = _H.at(2,0) * _point.x + _H.at(2,1) * _point.y + _H.at(2,2);
60 | if ( s != 0 )
61 | {
62 | ret.x = ( u / s );
63 | ret.y = ( v / s );
64 | }
65 | return ret;
66 | }
67 | Point3d IPM::applyHomography( const Point3d& _point )
68 | {
69 | return applyHomography( _point, m_H );
70 | }
71 | Point3d IPM::applyHomographyInv( const Point3d& _point )
72 | {
73 | return applyHomography( _point, m_H_inv );
74 | }
75 | Point3d IPM::applyHomography( const Point3d& _point, const cv::Mat& _H )
76 | {
77 | Point3d ret = Point3d( -1, -1, 1 );
78 |
79 | const double u = _H.at(0,0) * _point.x + _H.at(0,1) * _point.y + _H.at(0,2) * _point.z;
80 | const double v = _H.at(1,0) * _point.x + _H.at(1,1) * _point.y + _H.at(1,2) * _point.z;
81 | const double s = _H.at(2,0) * _point.x + _H.at(2,1) * _point.y + _H.at(2,2) * _point.z;
82 | if ( s != 0 )
83 | {
84 | ret.x = ( u / s );
85 | ret.y = ( v / s );
86 | }
87 | else
88 | ret.z = 0;
89 | return ret;
90 | }
91 |
92 | // Private
93 | void IPM::createMaps()
94 | {
95 | // Create remap images
96 | m_mapX.create(m_dstSize, CV_32F);
97 | m_mapY.create(m_dstSize, CV_32F);
98 | //#pragma omp parallel for schedule(dynamic)
99 | for( int j = 0; j < m_dstSize.height; ++j )
100 | {
101 | float* ptRowX = m_mapX.ptr(j);
102 | float* ptRowY = m_mapY.ptr(j);
103 | //#pragma omp parallel for schedule(dynamic)
104 | for( int i = 0; i < m_dstSize.width; ++i )
105 | {
106 | Point2f pt = applyHomography( Point2f( static_cast(i), static_cast(j) ), m_H_inv );
107 | ptRowX[i] = pt.x;
108 | ptRowY[i] = pt.y;
109 | }
110 | }
111 |
112 | m_invMapX.create(m_origSize, CV_32F);
113 | m_invMapY.create(m_origSize, CV_32F);
114 |
115 | //#pragma omp parallel for schedule(dynamic)
116 | for( int j = 0; j < m_origSize.height; ++j )
117 | {
118 | float* ptRowX = m_invMapX.ptr(j);
119 | float* ptRowY = m_invMapY.ptr(j);
120 | //#pragma omp parallel for schedule(dynamic)
121 | for( int i = 0; i < m_origSize.width; ++i )
122 | {
123 | Point2f pt = applyHomography( Point2f( static_cast(i), static_cast(j) ), m_H );
124 | ptRowX[i] = pt.x;
125 | ptRowY[i] = pt.y;
126 | }
127 | }
128 | }
129 |
--------------------------------------------------------------------------------
/windows/src/IPM.h:
--------------------------------------------------------------------------------
1 | #ifndef __IPM_H__
2 | #define __IPM_H__
3 |
4 | #include
5 | #include
6 | #include
7 | #include