├── LICENSE.txt
├── README.md
├── data
├── README.md
├── correct_coordinates.csv
├── correct_train.csv
└── train.csv
├── dataset.py
├── inference.py
├── models
├── __init__.py
├── model_cnet.py
└── model_countception.py
├── mytransforms.py
├── scripts
├── other
│ ├── README.md
│ ├── __notebook__.ipynb
│ └── script.py
└── preprocess.py
├── setup.py
├── train.py
├── utils.py
└── utils_cython.pyx
/LICENSE.txt:
--------------------------------------------------------------------------------
1 |
2 | Apache License
3 | Version 2.0, January 2004
4 | http://www.apache.org/licenses/
5 |
6 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
7 |
8 | 1. Definitions.
9 |
10 | "License" shall mean the terms and conditions for use, reproduction,
11 | and distribution as defined by Sections 1 through 9 of this document.
12 |
13 | "Licensor" shall mean the copyright owner or entity authorized by
14 | the copyright owner that is granting the License.
15 |
16 | "Legal Entity" shall mean the union of the acting entity and all
17 | other entities that control, are controlled by, or are under common
18 | control with that entity. For the purposes of this definition,
19 | "control" means (i) the power, direct or indirect, to cause the
20 | direction or management of such entity, whether by contract or
21 | otherwise, or (ii) ownership of fifty percent (50%) or more of the
22 | outstanding shares, or (iii) beneficial ownership of such entity.
23 |
24 | "You" (or "Your") shall mean an individual or Legal Entity
25 | exercising permissions granted by this License.
26 |
27 | "Source" form shall mean the preferred form for making modifications,
28 | including but not limited to software source code, documentation
29 | source, and configuration files.
30 |
31 | "Object" form shall mean any form resulting from mechanical
32 | transformation or translation of a Source form, including but
33 | not limited to compiled object code, generated documentation,
34 | and conversions to other media types.
35 |
36 | "Work" shall mean the work of authorship, whether in Source or
37 | Object form, made available under the License, as indicated by a
38 | copyright notice that is included in or attached to the work
39 | (an example is provided in the Appendix below).
40 |
41 | "Derivative Works" shall mean any work, whether in Source or Object
42 | form, that is based on (or derived from) the Work and for which the
43 | editorial revisions, annotations, elaborations, or other modifications
44 | represent, as a whole, an original work of authorship. For the purposes
45 | of this License, Derivative Works shall not include works that remain
46 | separable from, or merely link (or bind by name) to the interfaces of,
47 | the Work and Derivative Works thereof.
48 |
49 | "Contribution" shall mean any work of authorship, including
50 | the original version of the Work and any modifications or additions
51 | to that Work or Derivative Works thereof, that is intentionally
52 | submitted to Licensor for inclusion in the Work by the copyright owner
53 | or by an individual or Legal Entity authorized to submit on behalf of
54 | the copyright owner. For the purposes of this definition, "submitted"
55 | means any form of electronic, verbal, or written communication sent
56 | to the Licensor or its representatives, including but not limited to
57 | communication on electronic mailing lists, source code control systems,
58 | and issue tracking systems that are managed by, or on behalf of, the
59 | Licensor for the purpose of discussing and improving the Work, but
60 | excluding communication that is conspicuously marked or otherwise
61 | designated in writing by the copyright owner as "Not a Contribution."
62 |
63 | "Contributor" shall mean Licensor and any individual or Legal Entity
64 | on behalf of whom a Contribution has been received by Licensor and
65 | subsequently incorporated within the Work.
66 |
67 | 2. Grant of Copyright License. Subject to the terms and conditions of
68 | this License, each Contributor hereby grants to You a perpetual,
69 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable
70 | copyright license to reproduce, prepare Derivative Works of,
71 | publicly display, publicly perform, sublicense, and distribute the
72 | Work and such Derivative Works in Source or Object form.
73 |
74 | 3. Grant of Patent License. Subject to the terms and conditions of
75 | this License, each Contributor hereby grants to You a perpetual,
76 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable
77 | (except as stated in this section) patent license to make, have made,
78 | use, offer to sell, sell, import, and otherwise transfer the Work,
79 | where such license applies only to those patent claims licensable
80 | by such Contributor that are necessarily infringed by their
81 | Contribution(s) alone or by combination of their Contribution(s)
82 | with the Work to which such Contribution(s) was submitted. If You
83 | institute patent litigation against any entity (including a
84 | cross-claim or counterclaim in a lawsuit) alleging that the Work
85 | or a Contribution incorporated within the Work constitutes direct
86 | or contributory patent infringement, then any patent licenses
87 | granted to You under this License for that Work shall terminate
88 | as of the date such litigation is filed.
89 |
90 | 4. Redistribution. You may reproduce and distribute copies of the
91 | Work or Derivative Works thereof in any medium, with or without
92 | modifications, and in Source or Object form, provided that You
93 | meet the following conditions:
94 |
95 | (a) You must give any other recipients of the Work or
96 | Derivative Works a copy of this License; and
97 |
98 | (b) You must cause any modified files to carry prominent notices
99 | stating that You changed the files; and
100 |
101 | (c) You must retain, in the Source form of any Derivative Works
102 | that You distribute, all copyright, patent, trademark, and
103 | attribution notices from the Source form of the Work,
104 | excluding those notices that do not pertain to any part of
105 | the Derivative Works; and
106 |
107 | (d) If the Work includes a "NOTICE" text file as part of its
108 | distribution, then any Derivative Works that You distribute must
109 | include a readable copy of the attribution notices contained
110 | within such NOTICE file, excluding those notices that do not
111 | pertain to any part of the Derivative Works, in at least one
112 | of the following places: within a NOTICE text file distributed
113 | as part of the Derivative Works; within the Source form or
114 | documentation, if provided along with the Derivative Works; or,
115 | within a display generated by the Derivative Works, if and
116 | wherever such third-party notices normally appear. The contents
117 | of the NOTICE file are for informational purposes only and
118 | do not modify the License. You may add Your own attribution
119 | notices within Derivative Works that You distribute, alongside
120 | or as an addendum to the NOTICE text from the Work, provided
121 | that such additional attribution notices cannot be construed
122 | as modifying the License.
123 |
124 | You may add Your own copyright statement to Your modifications and
125 | may provide additional or different license terms and conditions
126 | for use, reproduction, or distribution of Your modifications, or
127 | for any such Derivative Works as a whole, provided Your use,
128 | reproduction, and distribution of the Work otherwise complies with
129 | the conditions stated in this License.
130 |
131 | 5. Submission of Contributions. Unless You explicitly state otherwise,
132 | any Contribution intentionally submitted for inclusion in the Work
133 | by You to the Licensor shall be under the terms and conditions of
134 | this License, without any additional terms or conditions.
135 | Notwithstanding the above, nothing herein shall supersede or modify
136 | the terms of any separate license agreement you may have executed
137 | with Licensor regarding such Contributions.
138 |
139 | 6. Trademarks. This License does not grant permission to use the trade
140 | names, trademarks, service marks, or product names of the Licensor,
141 | except as required for reasonable and customary use in describing the
142 | origin of the Work and reproducing the content of the NOTICE file.
143 |
144 | 7. Disclaimer of Warranty. Unless required by applicable law or
145 | agreed to in writing, Licensor provides the Work (and each
146 | Contributor provides its Contributions) on an "AS IS" BASIS,
147 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
148 | implied, including, without limitation, any warranties or conditions
149 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
150 | PARTICULAR PURPOSE. You are solely responsible for determining the
151 | appropriateness of using or redistributing the Work and assume any
152 | risks associated with Your exercise of permissions under this License.
153 |
154 | 8. Limitation of Liability. In no event and under no legal theory,
155 | whether in tort (including negligence), contract, or otherwise,
156 | unless required by applicable law (such as deliberate and grossly
157 | negligent acts) or agreed to in writing, shall any Contributor be
158 | liable to You for damages, including any direct, indirect, special,
159 | incidental, or consequential damages of any character arising as a
160 | result of this License or out of the use or inability to use the
161 | Work (including but not limited to damages for loss of goodwill,
162 | work stoppage, computer failure or malfunction, or any and all
163 | other commercial damages or losses), even if such Contributor
164 | has been advised of the possibility of such damages.
165 |
166 | 9. Accepting Warranty or Additional Liability. While redistributing
167 | the Work or Derivative Works thereof, You may choose to offer,
168 | and charge a fee for, acceptance of support, warranty, indemnity,
169 | or other liability obligations and/or rights consistent with this
170 | License. However, in accepting such obligations, You may act only
171 | on Your own behalf and on Your sole responsibility, not on behalf
172 | of any other Contributor, and only if You agree to indemnify,
173 | defend, and hold each Contributor harmless for any liability
174 | incurred by, or claims asserted against, such Contributor by reason
175 | of your accepting any such warranty or additional liability.
176 |
177 | END OF TERMS AND CONDITIONS
178 |
179 | APPENDIX: How to apply the Apache License to your work.
180 |
181 | To apply the Apache License to your work, attach the following
182 | boilerplate notice, with the fields enclosed by brackets "[]"
183 | replaced with your own identifying information. (Don't include
184 | the brackets!) The text should be enclosed in the appropriate
185 | comment syntax for the file format. We also recommend that a
186 | file or class name and description of purpose be included on the
187 | same "printed page" as the copyright notice for easier
188 | identification within third-party archives.
189 |
190 | Copyright [yyyy] [name of copyright owner]
191 |
192 | Licensed under the Apache License, Version 2.0 (the "License");
193 | you may not use this file except in compliance with the License.
194 | You may obtain a copy of the License at
195 |
196 | http://www.apache.org/licenses/LICENSE-2.0
197 |
198 | Unless required by applicable law or agreed to in writing, software
199 | distributed under the License is distributed on an "AS IS" BASIS,
200 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
201 | See the License for the specific language governing permissions and
202 | limitations under the License.
203 |
--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | # Pytorch Counting Models for Kaggle Sea Lion Count Challenge
2 |
3 | ## Overview
4 | With less than two weeks remaining, I decided to jump into the Kaggle Sea Lion count competition (https://www.kaggle.com/c/noaa-fisheries-steller-sea-lion-population-count) and see if I could get results implementing a few CNN based counting models I'd been reading about.
5 |
6 | Basically an excuse to try Pytorch and experiment with some new models. Most of my other NN hacking has been in Tensorflow, Torch, or Theano.
7 |
8 | As far as the competition is concerned, these models were a fail. I'm not convinced they couldn't work but I didn't have time to find appropriate hyper parameters, tweak the models, or fix issues in my code to produce anything reasonable.
9 |
10 |
11 | I implemented two models
12 | * Count-ception -- From (https://arxiv.org/abs/1703.08710)
13 | * Count-net -- My own mashup of (https://arxiv.org/pdf/1705.10118.pdf) and (https://www.robots.ox.ac.uk/~vgg/publications/2015/Xie15/weidi15.pdf)
14 |
15 | I wanted to give the FCRN described in https://www.robots.ox.ac.uk/~vgg/publications/2015/Xie15/weidi15.pdf a shot, but the layer description in the paper between conv4 and FC was vague and I ran out of time.
16 |
17 | What's working:
18 | * Kaggle Sealion patch based data processing pipeline with augmentation
19 | * Density or redundant count-ception target generation
20 | * Model training (loss curve looks reasonable)
21 | * Inference (submission generation)
22 |
23 | What's not:
24 | * Good results. Both models train but the appropriate features for sea lion counting and category discrimination do not appear to be learned. Counts are way off. Doing regression across multiple categories of similar looking objects is likely making this a very challenging objective.
25 | * Validation
26 |
27 |
28 | ## Examples
29 |
30 | Train:
31 |
32 | python train.py /data/sealion/Train-processed/ --batch-size 8 --num-processes 4 --num-gpu 2 --lr 0.001 --opt adadelta --model cc --loss l1
33 |
34 | Inference:
35 |
36 | python inference.py /data/x/sealion/Test/ --batch-size 8 --num-processes 4 --restore-checkpoint output/train/20170625-200215/checkpoint-1.pth.tar
37 |
38 |
39 | Build 'utils_cython' module for overlapping patch merge:
40 |
41 | setup.py build_ext --inplace
42 |
43 |
--------------------------------------------------------------------------------
/data/README.md:
--------------------------------------------------------------------------------
1 | Corrected coordinates and sealion counts (correct_coordinates.csv and correct_train.csv) sourced from:
2 | * https://github.com/LivingProgram/kaggle-sea-lion-data
3 | * https://www.kaggle.com/c/noaa-fisheries-steller-sea-lion-population-count/discussion/32857
4 |
5 | Thanks!
--------------------------------------------------------------------------------
/data/correct_train.csv:
--------------------------------------------------------------------------------
1 | train_id,adult_males,subadult_males,adult_females,juveniles,pups
2 | 0,62,12,486,42,344
3 | 1,2,20,0,12,0
4 | 2,2,0,37,19,0
5 | 4,6,9,2,0,0
6 | 5,6,4,14,4,19
7 | 6,2,5,20,18,0
8 | 8,9,5,76,4,51
9 | 10,5,4,7,5,0
10 | 11,3,5,36,13,0
11 | 12,8,9,13,1,7
12 | 13,1,5,20,13,0
13 | 14,1,1,0,27,0
14 | 15,2,3,33,56,0
15 | 16,1,1,32,35,0
16 | 17,4,4,60,18,0
17 | 18,2,3,0,0,0
18 | 19,7,3,34,24,1
19 | 20,7,21,31,4,1
20 | 22,1,1,0,5,0
21 | 23,1,0,1,6,0
22 | 24,4,2,0,0,0
23 | 25,1,2,0,0,0
24 | 26,0,8,0,4,0
25 | 27,1,4,0,8,0
26 | 28,2,0,0,0,0
27 | 29,4,4,44,15,0
28 | 31,3,2,20,76,1
29 | 32,18,6,159,18,111
30 | 33,0,1,18,8,0
31 | 35,4,1,0,0,0
32 | 36,8,17,0,0,0
33 | 37,1,2,5,0,0
34 | 38,3,0,33,0,0
35 | 39,10,14,62,10,15
36 | 40,2,2,62,7,0
37 | 41,15,0,85,18,59
38 | 42,7,4,10,1,0
39 | 43,28,4,338,47,189
40 | 44,3,2,25,15,0
41 | 45,4,7,100,27,0
42 | 46,1,4,0,0,0
43 | 47,13,14,48,3,33
44 | 48,5,10,66,24,0
45 | 49,0,0,4,15,0
46 | 50,1,0,0,0,0
47 | 51,9,13,10,35,0
48 | 52,2,3,20,23,0
49 | 53,2,1,15,6,0
50 | 54,2,5,22,26,0
51 | 55,6,2,59,74,0
52 | 56,9,3,91,5,93
53 | 57,2,1,0,0,0
54 | 58,38,17,143,71,145
55 | 59,6,4,16,55,1
56 | 60,5,4,0,0,0
57 | 61,0,0,1,0,0
58 | 62,6,2,221,28,1
59 | 63,7,22,58,15,59
60 | 64,5,5,6,0,0
61 | 65,29,8,267,14,270
62 | 66,9,5,23,17,2
63 | 67,9,3,30,4,25
64 | 68,4,3,3,0,0
65 | 69,4,2,46,31,14
66 | 70,9,11,15,636,0
67 | 72,13,3,56,37,20
68 | 73,3,0,20,6,7
69 | 74,1,2,0,4,0
70 | 75,3,1,0,34,0
71 | 76,5,3,0,0,0
72 | 77,19,11,136,37,83
73 | 78,2,2,6,0,5
74 | 79,11,5,37,0,26
75 | 80,0,12,0,19,0
76 | 82,0,0,37,17,0
77 | 83,5,2,45,41,0
78 | 84,1,1,0,12,0
79 | 85,0,1,0,0,0
80 | 86,2,2,4,2,2
81 | 87,7,1,43,16,0
82 | 88,1,2,1,24,0
83 | 90,1,0,17,12,0
84 | 91,17,0,229,1,156
85 | 92,3,5,1,1,1
86 | 93,23,21,144,100,62
87 | 94,4,4,17,0,0
88 | 95,5,0,19,58,0
89 | 96,9,3,167,5,0
90 | 98,0,6,0,0,0
91 | 99,3,1,11,68,0
92 | 100,12,3,50,4,27
93 | 101,10,0,85,56,0
94 | 102,13,0,92,3,73
95 | 103,3,0,0,0,0
96 | 104,3,3,38,43,0
97 | 105,2,0,32,13,3
98 | 106,0,4,0,5,0
99 | 107,7,11,3,1,2
100 | 108,3,4,0,3,0
101 | 109,4,2,35,4,27
102 | 110,16,8,95,13,79
103 | 111,23,9,123,40,80
104 | 112,11,20,387,190,0
105 | 113,0,9,10,3,0
106 | 114,5,7,12,0,9
107 | 115,0,13,0,4,0
108 | 116,0,0,4,9,0
109 | 117,0,1,1,0,0
110 | 118,1,1,10,43,0
111 | 119,7,7,19,2,11
112 | 120,2,0,5,1,3
113 | 121,6,1,90,7,67
114 | 122,11,43,37,14,0
115 | 123,0,0,1,0,0
116 | 124,19,8,104,84,46
117 | 125,6,1,36,29,8
118 | 126,35,14,127,7,117
119 | 127,1,9,15,64,0
120 | 128,11,22,53,43,0
121 | 129,0,3,0,36,0
122 | 130,13,30,43,17,0
123 | 131,5,11,25,73,2
124 | 132,2,2,0,0,0
125 | 133,1,1,2,2,0
126 | 134,3,2,14,6,0
127 | 135,1,0,10,4,0
128 | 136,1,9,3,21,0
129 | 137,6,3,0,0,0
130 | 138,1,0,25,12,0
131 | 139,2,4,9,0,8
132 | 140,1,2,0,0,0
133 | 141,5,28,0,16,0
134 | 142,1,1,3,5,0
135 | 143,0,1,0,0,0
136 | 144,8,0,18,0,20
137 | 145,4,9,60,70,0
138 | 146,4,1,24,23,5
139 | 147,6,4,44,4,26
140 | 148,0,3,0,5,0
141 | 149,2,5,0,2,0
142 | 150,0,2,0,0,0
143 | 152,5,4,0,0,0
144 | 153,12,27,0,0,0
145 | 154,6,13,62,25,0
146 | 155,23,11,114,5,64
147 | 156,2,1,11,0,0
148 | 157,1,1,15,0,0
149 | 158,20,0,90,3,87
150 | 159,0,6,0,0,0
151 | 160,12,5,199,65,30
152 | 161,2,5,5,0,0
153 | 162,4,1,58,37,0
154 | 163,11,8,89,15,72
155 | 164,32,15,248,14,169
156 | 165,2,5,4,3,0
157 | 166,5,5,56,25,0
158 | 167,1,0,0,0,0
159 | 168,6,4,5,101,0
160 | 169,4,2,35,14,11
161 | 170,10,0,82,4,67
162 | 171,0,23,3,37,0
163 | 172,2,0,6,23,0
164 | 173,4,0,6,6,0
165 | 174,4,0,35,0,29
166 | 175,2,0,4,17,0
167 | 176,17,9,92,6,56
168 | 177,2,10,0,8,0
169 | 178,2,0,1,0,0
170 | 179,0,5,0,1,0
171 | 180,3,1,7,4,0
172 | 181,6,6,48,37,2
173 | 182,1,2,13,10,0
174 | 183,6,7,31,28,0
175 | 185,8,1,13,0,8
176 | 186,3,4,112,146,0
177 | 187,54,14,592,8,434
178 | 188,1,3,10,19,0
179 | 189,64,40,382,23,135
180 | 190,1,0,2,6,0
181 | 191,1,5,0,0,0
182 | 192,1,0,0,0,0
183 | 193,2,1,29,15,0
184 | 194,1,0,0,0,0
185 | 195,4,0,63,24,0
186 | 196,12,8,106,11,64
187 | 197,7,3,80,6,62
188 | 198,3,2,0,0,0
189 | 199,1,0,0,0,0
190 | 200,36,3,56,2,5
191 | 201,9,2,44,13,33
192 | 202,3,0,11,23,0
193 | 203,8,8,43,59,2
194 | 204,4,8,1,6,1
195 | 205,14,1,135,10,68
196 | 206,0,5,1,19,1
197 | 207,0,2,0,2,0
198 | 208,4,3,0,0,0
199 | 209,4,9,0,1,0
200 | 210,2,9,1,15,0
201 | 211,19,4,185,20,243
202 | 212,4,6,16,0,0
203 | 213,11,12,26,44,6
204 | 214,4,0,18,5,5
205 | 216,5,0,28,0,33
206 | 217,16,4,140,12,142
207 | 218,0,2,0,1,0
208 | 219,11,7,37,39,0
209 | 220,0,1,0,0,0
210 | 221,6,1,26,9,2
211 | 222,2,0,0,0,0
212 | 223,0,0,1,0,0
213 | 224,12,1,58,3,68
214 | 225,0,1,22,6,0
215 | 226,17,8,207,3,144
216 | 227,5,1,52,111,0
217 | 228,1,9,61,38,0
218 | 229,1,1,0,1,0
219 | 230,0,0,16,5,0
220 | 231,7,4,7,1,1
221 | 232,9,2,21,1,16
222 | 233,1,0,0,0,0
223 | 235,29,3,277,99,135
224 | 236,3,0,29,44,0
225 | 237,0,1,29,38,0
226 | 238,3,3,8,1,3
227 | 239,4,10,8,17,0
228 | 240,25,8,161,23,114
229 | 241,3,0,101,54,0
230 | 243,4,0,0,0,0
231 | 244,1,1,3,13,0
232 | 245,7,2,74,29,8
233 | 246,5,3,24,31,1
234 | 247,3,1,34,18,0
235 | 248,4,0,64,34,0
236 | 249,4,8,13,38,0
237 | 250,2,4,2,0,0
238 | 251,1,0,5,31,0
239 | 252,21,2,105,15,67
240 | 253,2,0,0,0,0
241 | 254,2,9,7,1,6
242 | 255,0,3,0,0,0
243 | 256,15,0,92,53,59
244 | 257,1,11,23,53,9
245 | 258,19,11,350,265,1
246 | 259,3,3,32,20,0
247 | 260,3,1,8,51,0
248 | 261,3,4,29,9,2
249 | 262,3,0,34,31,5
250 | 263,12,2,57,32,23
251 | 264,0,0,3,3,0
252 | 265,1,5,0,0,0
253 | 266,4,0,39,35,0
254 | 267,14,8,62,14,44
255 | 269,3,2,0,0,0
256 | 270,3,0,21,29,0
257 | 271,30,7,288,34,217
258 | 272,0,2,0,4,0
259 | 273,14,3,296,7,250
260 | 274,1,0,0,0,0
261 | 275,0,1,25,40,0
262 | 276,6,3,37,0,0
263 | 277,5,6,68,102,0
264 | 278,5,0,18,19,2
265 | 279,1,0,8,4,0
266 | 280,12,5,19,3,29
267 | 281,3,3,51,20,0
268 | 282,0,3,3,0,0
269 | 283,4,7,26,12,0
270 | 284,2,0,16,15,0
271 | 285,0,2,0,0,0
272 | 286,2,2,0,0,0
273 | 287,10,8,1,0,0
274 | 288,4,0,12,15,1
275 | 289,0,2,38,15,0
276 | 291,39,20,275,23,237
277 | 292,5,5,49,42,1
278 | 293,6,4,38,2,31
279 | 294,1,1,2,5,0
280 | 295,16,6,51,29,26
281 | 296,1,6,0,2,0
282 | 297,3,36,24,39,0
283 | 298,22,10,172,18,183
284 | 299,27,9,209,32,55
285 | 300,6,3,7,11,0
286 | 301,6,3,67,5,35
287 | 302,1,0,0,34,0
288 | 303,1,2,0,25,0
289 | 304,2,3,0,82,0
290 | 305,1,2,3,0,0
291 | 306,14,4,71,45,17
292 | 307,8,0,17,6,17
293 | 308,0,2,4,2,0
294 | 309,5,0,61,23,8
295 | 310,4,8,95,30,0
296 | 312,1,1,21,14,0
297 | 313,5,3,83,30,0
298 | 314,1,2,0,0,0
299 | 315,1,2,0,0,0
300 | 316,5,0,0,1,0
301 | 317,1,2,34,16,0
302 | 318,2,9,33,42,1
303 | 319,17,12,145,68,49
304 | 320,16,4,227,39,11
305 | 321,2,3,0,1,0
306 | 322,8,3,63,22,10
307 | 323,3,0,1,0,1
308 | 324,2,3,12,33,0
309 | 325,0,2,2,79,0
310 | 326,1,3,69,80,0
311 | 327,0,2,4,0,0
312 | 328,6,4,18,4,13
313 | 329,18,3,72,6,49
314 | 330,1,0,25,66,0
315 | 332,10,4,57,27,21
316 | 333,0,15,0,74,0
317 | 334,9,14,21,39,9
318 | 335,6,36,18,12,0
319 | 336,10,2,77,18,27
320 | 337,0,3,4,86,0
321 | 338,37,7,275,41,248
322 | 339,2,2,5,0,0
323 | 340,21,4,181,23,145
324 | 341,0,1,0,1,0
325 | 342,3,0,21,16,0
326 | 343,2,1,4,2,3
327 | 345,1,0,16,2,0
328 | 346,7,1,46,39,0
329 | 347,10,55,2,23,0
330 | 348,0,2,0,17,0
331 | 349,24,29,21,11,0
332 | 350,4,13,0,0,0
333 | 351,1,0,0,0,0
334 | 352,2,1,0,0,0
335 | 353,5,4,78,29,0
336 | 354,2,1,31,4,0
337 | 355,0,4,0,1,0
338 | 356,0,0,1,0,0
339 | 357,4,1,0,0,0
340 | 358,0,3,0,38,0
341 | 359,10,4,51,12,46
342 | 360,2,4,23,77,0
343 | 361,27,1,146,38,54
344 | 362,18,1,127,4,97
345 | 363,1,7,4,53,1
346 | 364,6,0,24,0,45
347 | 365,3,6,0,0,0
348 | 366,1,1,2,2,0
349 | 367,15,4,99,4,55
350 | 368,6,0,37,3,34
351 | 369,8,41,0,24,0
352 | 370,9,4,43,7,16
353 | 371,15,2,62,5,49
354 | 372,29,6,136,33,82
355 | 373,12,6,69,54,6
356 | 374,0,0,36,43,0
357 | 375,3,0,0,0,0
358 | 376,2,3,24,4,0
359 | 377,8,5,27,2,29
360 | 378,5,1,43,1,0
361 | 379,4,7,23,45,5
362 | 381,1,0,0,0,0
363 | 382,1,0,12,22,0
364 | 383,8,8,70,124,0
365 | 385,2,2,0,1,0
366 | 386,4,1,15,8,10
367 | 387,2,0,0,0,0
368 | 388,1,0,12,24,0
369 | 389,0,1,42,14,0
370 | 390,4,5,4,2,3
371 | 391,7,3,8,0,8
372 | 392,2,0,15,19,0
373 | 393,7,6,33,2,21
374 | 394,1,2,0,0,0
375 | 395,2,0,25,15,0
376 | 396,1,11,0,2,0
377 | 397,2,1,2,0,0
378 | 398,0,4,0,11,0
379 | 399,6,13,30,70,0
380 | 400,3,1,7,0,0
381 | 401,8,8,33,32,16
382 | 402,6,11,300,71,0
383 | 403,1,17,22,195,2
384 | 404,2,0,12,1,6
385 | 405,1,2,4,7,0
386 | 407,5,2,4,0,0
387 | 408,9,5,45,2,34
388 | 409,4,0,13,6,1
389 | 410,1,2,0,0,0
390 | 411,17,0,329,345,0
391 | 412,5,7,104,61,0
392 | 413,8,2,0,0,0
393 | 414,3,7,10,13,0
394 | 415,4,0,20,2,0
395 | 416,2,34,7,340,0
396 | 417,6,2,4,0,0
397 | 418,10,6,140,150,0
398 | 419,3,1,0,0,0
399 | 420,8,6,13,4,16
400 | 422,10,0,93,19,90
401 | 423,0,4,0,4,0
402 | 424,1,0,34,18,0
403 | 425,0,1,14,16,0
404 | 426,2,6,11,42,5
405 | 427,1,3,4,28,0
406 | 428,1,0,1,1,0
407 | 429,3,18,19,25,0
408 | 430,4,0,10,6,1
409 | 431,0,2,5,8,0
410 | 432,4,8,23,23,0
411 | 433,0,3,5,43,0
412 | 434,7,3,21,101,0
413 | 435,0,4,1,19,0
414 | 436,2,1,0,0,0
415 | 437,30,1,218,4,192
416 | 438,15,6,169,0,123
417 | 439,2,3,21,16,0
418 | 440,2,13,4,7,0
419 | 441,1,0,8,15,0
420 | 442,1,2,5,23,0
421 | 443,4,1,12,35,0
422 | 444,1,4,0,0,0
423 | 445,1,0,11,16,0
424 | 446,9,1,62,19,32
425 | 447,0,4,0,2,0
426 | 448,1,1,2,1,1
427 | 449,1,6,0,0,0
428 | 450,1,0,0,0,0
429 | 451,2,3,44,29,0
430 | 452,35,7,207,9,100
431 | 453,2,1,9,20,0
432 | 454,2,5,22,35,1
433 | 455,3,0,49,17,2
434 | 456,3,3,5,0,0
435 | 457,2,1,5,13,0
436 | 458,1,9,7,6,0
437 | 459,15,7,58,3,30
438 | 460,1,1,22,20,0
439 | 461,1,2,17,40,0
440 | 462,14,6,60,6,58
441 | 463,6,0,57,7,7
442 | 464,0,1,0,16,0
443 | 465,10,2,94,171,0
444 | 466,3,1,9,3,7
445 | 467,1,4,0,0,0
446 | 468,8,14,39,10,0
447 | 470,0,0,1,1,0
448 | 471,4,2,7,4,0
449 | 472,6,4,45,15,0
450 | 473,2,0,19,60,0
451 | 474,1,2,7,26,0
452 | 476,9,0,42,39,0
453 | 477,9,5,50,16,14
454 | 478,8,9,49,13,44
455 | 479,5,4,0,0,0
456 | 480,3,2,30,9,1
457 | 481,3,2,4,45,0
458 | 482,7,1,44,6,33
459 | 483,1,2,3,30,0
460 | 484,6,40,0,5,0
461 | 485,10,6,45,35,0
462 | 486,3,2,0,1,1
463 | 487,3,0,0,0,0
464 | 488,6,11,145,78,0
465 | 489,2,0,2,16,0
466 | 491,0,1,9,4,0
467 | 492,2,1,9,21,1
468 | 493,0,21,0,65,0
469 | 494,24,44,72,19,53
470 | 495,2,0,20,2,20
471 | 496,0,2,5,0,4
472 | 497,1,2,0,0,0
473 | 498,0,1,1,19,0
474 | 500,2,6,0,1,0
475 | 501,2,0,2,3,0
476 | 502,6,3,59,32,0
477 | 503,0,5,6,31,0
478 | 504,2,0,0,3,0
479 | 505,1,8,0,41,0
480 | 506,0,0,9,21,0
481 | 508,0,0,4,1,0
482 | 509,23,10,86,63,43
483 | 510,5,1,0,0,0
484 | 511,3,4,16,27,0
485 | 512,1,4,0,0,0
486 | 513,7,0,8,2,7
487 | 514,6,3,260,109,0
488 | 515,0,8,2,1,0
489 | 516,13,3,92,14,61
490 | 517,2,1,4,0,0
491 | 518,4,7,59,30,0
492 | 519,6,3,44,17,13
493 | 520,6,2,21,0,0
494 | 521,1,0,7,5,0
495 | 522,5,3,67,43,5
496 | 523,16,4,64,13,16
497 | 524,4,2,1,1,0
498 | 525,3,0,9,35,0
499 | 526,4,10,20,19,13
500 | 527,2,0,29,6,1
501 | 528,8,0,30,4,24
502 | 529,5,2,15,13,0
503 | 532,2,0,0,0,0
504 | 533,2,2,0,2,0
505 | 534,0,1,0,0,0
506 | 535,7,0,81,36,7
507 | 536,15,1,52,4,44
508 | 537,1,4,2,8,0
509 | 538,10,2,163,9,115
510 | 539,1,5,0,0,0
511 | 540,22,21,40,11,1
512 | 541,4,1,12,2,0
513 | 542,3,2,0,0,0
514 | 543,3,2,0,6,0
515 | 544,4,6,8,2,0
516 | 545,3,14,36,21,0
517 | 546,3,1,16,47,0
518 | 547,2,2,18,34,0
519 | 548,3,0,5,4,0
520 | 549,1,4,31,0,0
521 | 550,2,0,11,14,0
522 | 551,1,0,13,13,0
523 | 552,2,3,52,46,0
524 | 553,5,3,3,1,0
525 | 554,1,8,9,1,1
526 | 555,3,3,20,28,0
527 | 556,13,17,133,32,102
528 | 557,0,3,0,3,0
529 | 558,14,3,87,21,39
530 | 559,2,2,29,41,0
531 | 560,1,2,0,38,0
532 | 561,3,0,33,42,0
533 | 562,3,4,31,68,0
534 | 563,6,5,68,6,88
535 | 564,1,0,0,0,0
536 | 565,0,0,6,8,0
537 | 566,2,2,0,0,0
538 | 567,5,3,66,30,1
539 | 568,4,0,9,17,2
540 | 569,19,9,113,24,105
541 | 570,2,3,0,0,0
542 | 571,0,0,30,13,0
543 | 572,0,0,11,9,0
544 | 573,2,5,0,0,0
545 | 574,4,2,21,9,7
546 | 575,2,1,19,3,0
547 | 576,3,12,24,24,0
548 | 577,3,1,116,97,0
549 | 578,6,3,26,0,16
550 | 579,0,0,24,33,0
551 | 580,3,5,0,0,0
552 | 581,21,10,130,15,16
553 | 582,1,4,0,0,0
554 | 583,2,2,3,1,0
555 | 584,7,0,4,0,3
556 | 585,2,1,4,16,0
557 | 586,3,0,1,5,1
558 | 587,31,13,200,75,268
559 | 588,5,31,13,11,0
560 | 589,1,1,21,30,0
561 | 590,47,16,322,41,300
562 | 591,22,14,134,116,43
563 | 592,1,1,0,0,0
564 | 593,1,2,32,58,0
565 | 594,0,1,1,0,0
566 | 595,2,0,28,14,0
567 | 596,5,4,30,47,0
568 | 597,0,0,1,0,0
569 | 598,2,1,32,15,0
570 | 599,3,8,1,5,0
571 | 600,0,1,19,27,0
572 | 601,9,4,38,92,0
573 | 602,3,17,0,0,0
574 | 603,2,3,1,0,0
575 | 604,12,4,105,204,2
576 | 606,15,5,288,11,211
577 | 608,2,0,12,24,0
578 | 609,1,5,0,0,0
579 | 610,5,5,17,12,0
580 | 611,2,1,32,13,0
581 | 612,3,0,1,0,0
582 | 613,4,6,15,2,1
583 | 615,2,1,35,22,0
584 | 616,8,18,79,25,0
585 | 617,14,12,88,47,38
586 | 618,5,23,98,25,0
587 | 619,6,2,27,5,12
588 | 620,5,6,20,1,21
589 | 622,6,2,32,1,38
590 | 623,6,0,0,0,0
591 | 624,7,6,38,1,20
592 | 625,28,5,149,11,129
593 | 626,1,0,0,0,0
594 | 627,1,1,0,0,0
595 | 628,11,4,18,1,0
596 | 629,10,9,75,1,57
597 | 630,2,0,0,0,0
598 | 631,2,2,0,0,0
599 | 632,4,3,1,6,0
600 | 633,5,1,39,15,23
601 | 634,1,0,0,0,0
602 | 635,3,1,20,7,0
603 | 636,2,2,34,2,0
604 | 637,6,18,85,17,40
605 | 639,4,12,31,81,2
606 | 640,8,3,33,8,0
607 | 641,1,2,70,102,0
608 | 642,1,9,0,0,0
609 | 643,1,5,0,21,0
610 | 645,10,7,73,50,67
611 | 646,1,1,11,3,0
612 | 647,1,2,0,1,0
613 | 648,5,0,56,0,41
614 | 649,2,0,19,49,0
615 | 650,2,25,0,13,0
616 | 651,1,7,11,26,0
617 | 652,0,3,0,0,0
618 | 653,2,3,0,18,0
619 | 654,1,3,0,8,0
620 | 655,5,3,21,1,17
621 | 656,4,6,14,7,9
622 | 657,0,1,0,0,0
623 | 658,5,8,0,0,0
624 | 659,1,1,21,25,0
625 | 660,1,0,1,4,0
626 | 661,2,3,17,49,0
627 | 662,8,6,245,20,0
628 | 663,1,4,9,11,0
629 | 664,18,15,129,67,71
630 | 665,0,9,0,2,0
631 | 666,0,4,4,14,0
632 | 667,4,1,1,0,1
633 | 668,0,3,4,32,0
634 | 669,5,4,16,0,0
635 | 670,0,4,21,19,0
636 | 671,1,1,11,58,0
637 | 672,5,25,0,10,0
638 | 673,1,0,0,16,0
639 | 674,20,3,112,23,118
640 | 675,4,1,5,0,3
641 | 676,16,10,43,6,21
642 | 677,7,6,6,4,2
643 | 678,3,7,19,45,0
644 | 679,15,1,94,16,83
645 | 680,17,5,160,8,162
646 | 681,2,1,24,9,0
647 | 682,1,2,6,9,0
648 | 683,1,2,3,2,0
649 | 684,17,4,89,45,54
650 | 685,6,12,0,0,0
651 | 686,3,34,0,59,0
652 | 688,0,8,0,13,0
653 | 689,1,0,1,0,0
654 | 690,5,5,34,7,28
655 | 691,1,1,0,0,0
656 | 692,0,0,15,45,0
657 | 693,3,10,7,2,0
658 | 694,3,1,5,1,0
659 | 695,1,0,2,0,2
660 | 696,2,4,0,2,0
661 | 697,1,1,2,0,0
662 | 698,1,0,0,14,0
663 | 699,18,13,39,2,14
664 | 700,9,4,81,7,23
665 | 701,12,10,91,19,158
666 | 702,8,4,25,10,10
667 | 703,7,0,38,52,0
668 | 704,26,6,167,25,105
669 | 705,3,0,19,38,2
670 | 706,2,4,39,18,0
671 | 707,4,21,1,7,0
672 | 708,3,7,0,3,0
673 | 709,1,2,19,50,0
674 | 710,6,3,0,0,0
675 | 711,4,7,5,3,0
676 | 713,2,3,10,5,0
677 | 714,13,2,85,7,21
678 | 715,9,1,57,7,35
679 | 716,5,5,53,30,0
680 | 717,1,4,16,2,1
681 | 718,13,3,39,1,42
682 | 719,1,3,15,10,0
683 | 720,8,1,51,7,24
684 | 722,19,6,164,7,109
685 | 723,8,0,41,5,45
686 | 724,6,39,4,85,0
687 | 725,6,2,68,100,0
688 | 726,3,4,3,19,0
689 | 727,13,6,103,18,66
690 | 728,13,4,104,11,77
691 | 729,1,4,3,28,0
692 | 730,1,3,0,0,0
693 | 731,4,0,12,0,8
694 | 732,4,6,0,0,0
695 | 733,1,0,0,0,0
696 | 734,1,1,0,0,0
697 | 735,0,3,9,0,0
698 | 736,9,36,0,2,0
699 | 737,2,3,0,0,0
700 | 738,1,1,0,8,0
701 | 739,4,3,1,0,1
702 | 740,9,4,20,8,2
703 | 741,33,5,228,38,176
704 | 742,6,1,23,6,14
705 | 743,12,5,113,13,91
706 | 744,6,3,51,497,0
707 | 745,32,10,269,30,183
708 | 746,1,0,11,19,0
709 | 747,17,6,251,17,225
710 | 748,0,5,0,0,0
711 | 749,6,0,24,8,2
712 | 750,3,1,9,5,0
713 | 751,0,4,0,0,0
714 | 752,6,1,16,6,12
715 | 753,8,1,35,19,18
716 | 754,2,0,19,29,0
717 | 755,5,17,51,139,0
718 | 756,0,0,20,34,0
719 | 757,1,0,0,0,0
720 | 758,3,0,12,85,0
721 | 759,18,2,144,13,119
722 | 760,3,15,0,7,0
723 | 761,9,19,53,58,0
724 | 762,2,2,2,0,0
725 | 763,0,1,0,72,0
726 | 764,4,4,26,2,6
727 | 765,0,1,20,19,0
728 | 766,1,1,1,0,0
729 | 768,2,1,0,0,0
730 | 769,12,0,161,15,130
731 | 770,1,4,24,5,0
732 | 771,7,1,113,28,0
733 | 772,0,1,0,1,0
734 | 773,28,8,213,5,138
735 | 774,20,0,139,12,114
736 | 775,3,4,51,4,49
737 | 776,8,2,25,2,29
738 | 777,2,3,0,0,0
739 | 778,2,3,0,1,0
740 | 780,0,0,3,3,0
741 | 782,0,3,95,22,0
742 | 783,17,8,88,18,10
743 | 784,11,5,211,211,0
744 | 785,41,8,288,22,230
745 | 786,1,2,0,0,0
746 | 787,3,4,0,10,0
747 | 788,3,18,12,5,0
748 | 789,3,2,23,1,18
749 | 790,7,2,40,4,36
750 | 791,4,4,37,3,26
751 | 792,10,0,49,1,42
752 | 793,11,5,8,0,0
753 | 795,3,0,7,0,0
754 | 796,1,0,1,8,0
755 | 797,1,0,0,0,0
756 | 798,2,1,0,0,0
757 | 799,4,12,13,74,3
758 | 801,9,3,56,30,57
759 | 802,11,6,48,54,12
760 | 803,21,10,259,40,151
761 | 804,6,1,56,52,0
762 | 805,12,1,60,14,31
763 | 806,5,11,14,3,10
764 | 807,7,0,57,7,56
765 | 808,1,0,0,0,0
766 | 809,3,0,48,43,1
767 | 810,15,4,68,28,20
768 | 812,3,0,16,3,0
769 | 813,1,3,6,20,0
770 | 814,1,6,0,5,0
771 | 815,5,3,2,4,0
772 | 816,3,0,35,33,0
773 | 817,4,5,26,39,0
774 | 818,1,0,4,7,0
775 | 819,2,0,0,0,0
776 | 820,4,3,44,4,41
777 | 821,1,0,3,2,0
778 | 822,30,10,305,11,210
779 | 823,2,7,29,12,0
780 | 824,2,4,0,0,0
781 | 825,3,1,0,1,0
782 | 826,0,2,3,0,0
783 | 827,4,7,4,0,0
784 | 828,15,9,48,4,35
785 | 829,4,3,15,9,8
786 | 830,2,10,4,32,0
787 | 831,3,2,6,1,0
788 | 832,0,4,0,18,0
789 | 833,22,3,226,43,215
790 | 834,0,7,0,1,0
791 | 835,4,6,1,2,0
792 | 836,2,3,3,1,0
793 | 837,1,0,0,0,0
794 | 838,11,28,2,5,0
795 | 841,3,6,0,26,0
796 | 842,2,4,6,4,0
797 | 843,3,0,6,4,0
798 | 844,9,7,5,4,0
799 | 845,8,1,9,17,0
800 | 846,1,1,13,3,0
801 | 847,27,3,227,46,151
802 | 848,1,1,5,23,0
803 | 849,6,1,24,5,10
804 | 850,6,13,13,0,0
805 | 851,1,2,10,8,0
806 | 852,6,27,3,10,3
807 | 853,0,4,0,0,0
808 | 854,1,2,0,1,0
809 | 855,3,2,24,78,0
810 | 856,8,4,14,5,0
811 | 857,0,0,0,0,0
812 | 858,4,12,20,67,0
813 | 859,3,6,8,33,0
814 | 860,3,2,20,11,7
815 | 861,4,6,110,0,0
816 | 862,4,2,18,99,1
817 | 863,1,1,0,0,0
818 | 864,14,2,207,20,172
819 | 865,0,0,0,1,0
820 | 866,2,0,4,14,0
821 | 867,0,2,5,11,0
822 | 868,5,11,62,94,0
823 | 870,2,0,7,0,0
824 | 871,23,10,193,43,172
825 | 872,7,11,132,36,0
826 | 873,1,0,30,76,0
827 | 874,18,7,118,47,117
828 | 875,4,8,60,14,0
829 | 876,1,0,0,0,0
830 | 877,1,3,7,3,0
831 | 878,1,0,16,36,0
832 | 879,0,0,28,75,0
833 | 880,2,0,9,23,0
834 | 881,44,16,173,58,117
835 | 883,1,3,0,0,0
836 | 884,3,0,59,13,7
837 | 885,3,0,21,12,0
838 | 886,2,0,30,0,15
839 | 887,1,2,0,0,0
840 | 888,1,0,5,4,0
841 | 889,8,0,32,4,14
842 | 890,3,0,10,3,0
843 | 891,7,0,27,5,22
844 | 892,16,6,144,3,124
845 | 893,1,3,30,35,8
846 | 894,2,2,41,39,0
847 | 895,3,2,52,15,7
848 | 896,4,0,0,0,0
849 | 897,0,4,21,11,0
850 | 898,10,4,79,6,45
851 | 899,2,2,3,3,0
852 | 900,42,5,681,26,529
853 | 902,0,2,0,0,0
854 | 904,1,1,0,0,0
855 | 906,4,1,18,7,2
856 | 907,3,3,32,55,0
857 | 908,4,2,0,0,0
858 | 910,3,16,0,1,0
859 | 911,4,5,149,50,0
860 | 912,30,2,248,13,205
861 | 914,3,23,15,115,0
862 | 915,0,0,2,5,0
863 | 916,1,3,12,13,0
864 | 917,10,8,31,109,0
865 | 918,4,2,22,7,3
866 | 919,5,3,41,64,3
867 | 920,11,2,280,212,2
868 | 921,1,0,25,52,0
869 | 922,4,4,7,6,0
870 | 923,8,2,106,11,86
871 | 924,23,3,131,8,124
872 | 925,3,2,18,59,0
873 | 926,2,0,3,31,0
874 | 928,1,0,0,22,0
875 | 929,3,0,23,50,0
876 | 930,8,0,36,8,21
877 | 931,3,2,20,17,0
878 | 932,5,8,21,67,0
879 | 933,6,7,68,69,21
880 | 934,4,2,36,16,0
881 | 935,0,1,0,0,0
882 | 936,1,0,1,2,0
883 | 937,4,9,26,10,0
884 | 938,0,0,4,0,0
885 | 939,16,3,100,8,66
886 | 940,2,4,3,74,0
887 | 941,6,2,46,1,37
888 | 942,1,6,0,37,0
889 | 943,1,1,24,26,0
890 | 944,1,4,1,0,0
891 | 945,6,0,79,6,45
892 | 947,6,4,29,6,24
893 |
--------------------------------------------------------------------------------
/data/train.csv:
--------------------------------------------------------------------------------
1 | train_id,adult_males,subadult_males,adult_females,juveniles,pups
2 | 0,62,12,486,42,344
3 | 1,2,20,0,12,0
4 | 2,2,0,38,20,0
5 | 3,8,5,41,7,38
6 | 4,6,9,2,0,0
7 | 5,6,4,14,4,19
8 | 6,2,5,20,18,0
9 | 7,6,2,33,16,3
10 | 8,9,5,76,4,51
11 | 9,4,1,64,36,11
12 | 10,5,4,7,5,0
13 | 11,3,7,36,13,0
14 | 12,8,9,13,1,7
15 | 13,2,5,20,13,0
16 | 14,1,1,0,27,0
17 | 15,2,3,35,61,0
18 | 16,1,1,32,35,0
19 | 17,4,4,60,18,0
20 | 18,2,4,0,0,0
21 | 19,7,3,34,24,1
22 | 20,7,21,31,4,1
23 | 21,6,1,25,9,0
24 | 22,1,1,0,5,0
25 | 23,1,0,1,6,0
26 | 24,4,2,0,0,0
27 | 25,1,2,0,0,0
28 | 26,0,8,0,4,0
29 | 27,1,4,0,8,0
30 | 28,2,0,0,0,0
31 | 29,4,4,44,15,0
32 | 30,2,0,1,3,0
33 | 31,3,2,20,76,1
34 | 32,18,6,159,18,111
35 | 33,0,1,18,8,0
36 | 34,4,1,127,237,0
37 | 35,4,1,0,0,0
38 | 36,9,19,0,0,0
39 | 37,1,2,5,0,0
40 | 38,3,0,36,2,0
41 | 39,10,14,62,10,15
42 | 40,2,2,63,8,0
43 | 41,15,0,85,18,59
44 | 42,7,4,10,1,0
45 | 43,28,4,338,47,189
46 | 44,3,2,25,15,0
47 | 45,4,7,100,27,0
48 | 46,1,4,0,0,0
49 | 47,13,16,48,3,33
50 | 48,5,10,66,24,0
51 | 49,0,0,4,15,0
52 | 50,1,0,0,0,0
53 | 51,9,13,10,35,0
54 | 52,3,3,20,23,0
55 | 53,2,1,15,6,0
56 | 54,2,5,22,26,0
57 | 55,6,2,59,74,0
58 | 56,9,3,91,5,93
59 | 57,2,1,0,0,0
60 | 58,36,17,143,71,145
61 | 59,6,4,16,55,1
62 | 60,5,4,0,0,0
63 | 61,0,0,1,0,0
64 | 62,6,2,222,28,1
65 | 63,7,22,58,15,60
66 | 64,5,5,6,0,0
67 | 65,29,8,267,14,270
68 | 66,0,0,0,0,0
69 | 67,10,7,30,4,26
70 | 68,4,3,3,0,0
71 | 69,4,2,46,31,14
72 | 70,9,11,15,636,0
73 | 71,3,2,21,18,0
74 | 72,13,3,56,37,20
75 | 73,3,0,20,6,8
76 | 74,1,2,0,4,0
77 | 75,3,1,0,34,0
78 | 76,5,3,0,0,0
79 | 77,20,11,136,37,84
80 | 78,2,2,6,0,5
81 | 79,11,5,37,0,26
82 | 80,0,12,0,19,0
83 | 81,1,3,15,0,0
84 | 82,0,0,37,17,0
85 | 83,5,2,44,44,0
86 | 84,1,1,0,12,0
87 | 85,0,1,0,0,0
88 | 86,2,2,4,2,2
89 | 87,7,1,43,16,0
90 | 88,1,2,1,24,0
91 | 89,1,1,1,0,0
92 | 90,1,0,17,12,0
93 | 91,17,0,229,1,157
94 | 92,3,5,1,1,1
95 | 93,24,21,144,99,62
96 | 94,4,4,17,0,0
97 | 95,5,0,19,58,0
98 | 96,9,3,167,5,0
99 | 97,0,0,19,11,0
100 | 98,0,6,0,0,0
101 | 99,3,1,13,68,0
102 | 100,12,3,50,4,27
103 | 101,10,0,85,56,0
104 | 102,13,0,92,3,73
105 | 103,3,0,0,0,0
106 | 104,3,3,38,43,0
107 | 105,2,0,33,17,3
108 | 106,0,4,0,5,0
109 | 107,7,11,3,1,2
110 | 108,4,4,0,3,0
111 | 109,4,2,35,4,27
112 | 110,16,8,95,13,80
113 | 111,23,9,123,40,80
114 | 112,11,20,387,190,0
115 | 113,0,9,10,3,0
116 | 114,5,7,12,0,9
117 | 115,0,13,0,4,0
118 | 116,0,0,4,9,0
119 | 117,0,1,1,0,0
120 | 118,1,1,10,43,0
121 | 119,7,7,19,2,11
122 | 120,2,0,5,1,3
123 | 121,6,1,90,7,67
124 | 122,11,43,40,15,0
125 | 123,0,0,1,0,0
126 | 124,19,8,104,84,46
127 | 125,6,1,36,29,8
128 | 126,35,14,127,7,117
129 | 127,1,10,15,65,0
130 | 128,11,22,53,43,0
131 | 129,0,3,0,43,0
132 | 130,13,30,43,17,0
133 | 131,5,11,25,73,2
134 | 132,2,2,0,0,0
135 | 133,1,1,2,2,0
136 | 134,3,2,15,7,0
137 | 135,1,0,10,4,0
138 | 136,1,9,3,22,0
139 | 137,6,3,0,0,0
140 | 138,1,0,25,12,0
141 | 139,2,4,9,0,8
142 | 140,1,2,0,0,0
143 | 141,5,28,0,16,0
144 | 142,1,1,3,5,0
145 | 143,0,1,0,0,0
146 | 144,8,0,18,0,20
147 | 145,4,9,60,70,0
148 | 146,4,1,25,23,5
149 | 147,6,4,44,4,26
150 | 148,0,3,0,6,0
151 | 149,2,5,0,2,0
152 | 150,0,2,0,0,0
153 | 151,6,4,21,0,16
154 | 152,5,4,0,0,0
155 | 153,12,27,0,0,0
156 | 154,6,13,62,25,0
157 | 155,23,11,114,5,65
158 | 156,2,1,11,0,0
159 | 157,1,1,15,0,0
160 | 158,20,0,90,3,87
161 | 159,0,6,0,0,0
162 | 160,12,5,199,65,30
163 | 161,2,5,5,0,0
164 | 162,4,1,58,37,0
165 | 163,11,8,89,15,72
166 | 164,32,15,248,14,169
167 | 165,2,5,4,3,0
168 | 166,5,5,56,25,0
169 | 167,1,0,0,0,0
170 | 168,6,4,5,101,0
171 | 169,4,2,35,14,11
172 | 170,10,0,82,4,68
173 | 171,0,23,3,37,0
174 | 172,2,0,6,23,0
175 | 173,4,0,6,6,0
176 | 174,4,0,35,0,29
177 | 175,2,1,4,18,0
178 | 176,17,9,92,6,56
179 | 177,2,10,0,8,0
180 | 178,2,0,2,2,0
181 | 179,2,5,0,1,0
182 | 180,3,1,7,4,0
183 | 181,6,6,49,38,2
184 | 182,1,2,13,10,0
185 | 183,6,7,31,28,0
186 | 184,6,16,19,7,0
187 | 185,8,1,13,0,8
188 | 186,3,5,112,146,0
189 | 187,56,14,592,8,434
190 | 188,1,3,10,19,0
191 | 189,64,40,382,23,135
192 | 190,1,0,2,6,0
193 | 191,1,5,0,0,0
194 | 192,1,0,0,0,0
195 | 193,2,1,29,15,0
196 | 194,1,0,0,0,0
197 | 195,4,0,63,24,0
198 | 196,12,8,106,11,64
199 | 197,7,3,79,6,62
200 | 198,3,2,0,0,0
201 | 199,1,0,0,0,0
202 | 200,36,3,56,2,5
203 | 201,9,2,44,13,33
204 | 202,3,0,11,23,0
205 | 203,8,8,43,59,2
206 | 204,4,8,1,6,1
207 | 205,14,1,135,10,68
208 | 206,0,5,1,19,1
209 | 207,1,1,3,4,0
210 | 208,4,3,0,0,0
211 | 209,4,9,0,1,0
212 | 210,2,9,1,15,0
213 | 211,19,5,185,20,243
214 | 212,4,6,16,0,0
215 | 213,11,12,26,44,6
216 | 214,6,0,18,5,5
217 | 215,0,4,0,0,0
218 | 216,6,0,29,0,33
219 | 217,16,4,140,12,142
220 | 218,0,2,1,1,1
221 | 219,11,7,37,39,0
222 | 220,0,1,0,0,0
223 | 221,5,1,26,9,2
224 | 222,2,0,0,0,0
225 | 223,0,0,1,0,0
226 | 224,12,1,58,3,68
227 | 225,0,1,22,6,0
228 | 226,17,8,207,3,144
229 | 227,5,1,52,111,0
230 | 228,1,9,61,38,0
231 | 229,1,1,0,1,0
232 | 230,0,0,16,5,0
233 | 231,7,4,7,1,1
234 | 232,9,2,21,1,16
235 | 233,1,0,0,0,0
236 | 234,0,2,46,25,0
237 | 235,31,3,278,99,135
238 | 236,3,0,29,44,0
239 | 237,0,1,29,38,0
240 | 238,3,3,8,1,3
241 | 239,4,10,8,17,0
242 | 240,25,11,160,23,114
243 | 241,3,0,101,54,0
244 | 242,1,0,0,0,0
245 | 243,4,0,0,0,0
246 | 244,1,1,3,13,0
247 | 245,7,2,74,29,8
248 | 246,5,3,24,31,1
249 | 247,3,1,34,18,0
250 | 248,4,0,64,34,0
251 | 249,4,8,13,38,0
252 | 250,2,4,2,0,0
253 | 251,1,0,5,31,0
254 | 252,23,4,105,15,67
255 | 253,2,0,0,0,0
256 | 254,2,9,7,1,6
257 | 255,0,3,0,0,0
258 | 256,15,0,92,54,59
259 | 257,1,11,23,53,9
260 | 258,19,11,350,266,1
261 | 259,3,3,32,20,0
262 | 260,3,1,8,51,0
263 | 261,3,4,29,9,2
264 | 262,3,0,34,31,5
265 | 263,12,2,57,32,23
266 | 264,0,0,3,3,0
267 | 265,1,8,0,2,0
268 | 266,4,0,39,35,0
269 | 267,14,8,62,14,44
270 | 268,3,8,10,1,0
271 | 269,3,2,0,0,0
272 | 270,3,0,21,29,0
273 | 271,30,7,289,34,217
274 | 272,0,2,0,4,0
275 | 273,14,3,296,7,250
276 | 274,1,0,0,0,0
277 | 275,0,1,25,40,0
278 | 276,6,3,37,0,0
279 | 277,5,6,68,102,0
280 | 278,5,0,18,19,2
281 | 279,1,0,8,4,0
282 | 280,12,5,19,3,29
283 | 281,3,3,51,20,0
284 | 282,0,3,3,0,0
285 | 283,4,7,26,12,0
286 | 284,2,0,16,15,0
287 | 285,0,2,0,0,0
288 | 286,2,2,0,0,0
289 | 287,10,8,1,0,0
290 | 288,4,0,12,15,1
291 | 289,0,2,38,15,0
292 | 290,2,0,0,0,0
293 | 291,39,20,275,23,237
294 | 292,2,3,23,15,1
295 | 293,6,4,38,2,33
296 | 294,1,1,2,5,0
297 | 295,16,6,51,29,26
298 | 296,1,6,0,2,0
299 | 297,3,40,24,39,0
300 | 298,22,10,172,18,184
301 | 299,27,10,209,32,55
302 | 300,6,3,7,11,0
303 | 301,6,3,67,5,35
304 | 302,1,0,0,34,0
305 | 303,1,2,0,24,0
306 | 304,2,3,0,82,0
307 | 305,1,2,3,0,0
308 | 306,14,4,71,45,17
309 | 307,8,0,17,6,17
310 | 308,0,2,4,2,0
311 | 309,5,0,61,23,9
312 | 310,4,8,95,30,0
313 | 311,2,1,14,9,3
314 | 312,1,1,22,14,0
315 | 313,5,3,83,30,0
316 | 314,1,2,0,0,0
317 | 315,1,2,0,0,0
318 | 316,5,0,0,1,0
319 | 317,1,2,34,16,0
320 | 318,2,9,33,42,1
321 | 319,17,12,145,68,49
322 | 320,16,4,227,39,11
323 | 321,2,3,0,1,0
324 | 322,8,3,63,22,10
325 | 323,4,0,1,0,1
326 | 324,2,3,12,33,0
327 | 325,0,2,2,80,0
328 | 326,1,3,69,80,0
329 | 327,0,2,4,0,0
330 | 328,5,4,18,5,13
331 | 329,18,3,72,6,49
332 | 330,1,0,25,67,0
333 | 331,6,18,10,0,0
334 | 332,10,4,57,27,21
335 | 333,0,15,0,74,0
336 | 334,9,14,21,39,9
337 | 335,6,36,19,12,0
338 | 336,10,2,77,18,27
339 | 337,0,3,4,86,0
340 | 338,36,7,275,41,249
341 | 339,2,2,5,0,0
342 | 340,21,4,181,23,145
343 | 341,0,1,0,1,0
344 | 342,3,0,24,16,0
345 | 343,2,1,4,2,3
346 | 344,2,0,18,10,0
347 | 345,1,0,16,2,0
348 | 346,7,1,46,39,0
349 | 347,10,55,2,23,0
350 | 348,0,2,0,17,0
351 | 349,24,29,21,11,0
352 | 350,4,13,0,0,0
353 | 351,2,0,0,0,0
354 | 352,2,1,0,0,0
355 | 353,5,4,78,29,0
356 | 354,2,1,31,4,0
357 | 355,0,4,0,1,0
358 | 356,0,0,1,0,0
359 | 357,4,1,0,0,0
360 | 358,0,3,0,38,0
361 | 359,10,4,54,12,46
362 | 360,2,4,23,77,0
363 | 361,27,1,146,38,54
364 | 362,18,1,127,5,99
365 | 363,1,7,4,53,1
366 | 364,6,0,24,0,45
367 | 365,4,7,0,0,0
368 | 366,1,1,2,2,0
369 | 367,15,4,99,4,55
370 | 368,6,0,37,3,34
371 | 369,8,43,0,24,0
372 | 370,9,4,43,7,16
373 | 371,15,2,62,5,49
374 | 372,29,6,136,33,82
375 | 373,12,6,69,54,6
376 | 374,0,0,36,43,0
377 | 375,4,0,0,0,0
378 | 376,2,3,24,4,0
379 | 377,8,5,27,2,29
380 | 378,5,1,43,1,0
381 | 379,4,7,23,45,5
382 | 380,2,0,83,31,0
383 | 381,1,0,0,0,0
384 | 382,1,0,12,22,0
385 | 383,8,10,70,124,0
386 | 384,7,5,61,19,0
387 | 385,2,2,0,1,0
388 | 386,4,3,15,8,10
389 | 387,2,0,0,0,0
390 | 388,2,0,12,24,0
391 | 389,0,1,42,14,0
392 | 390,4,5,4,2,3
393 | 391,7,3,8,0,8
394 | 392,2,0,15,19,0
395 | 393,7,6,33,2,21
396 | 394,1,6,0,0,0
397 | 395,2,1,25,16,0
398 | 396,1,11,0,2,0
399 | 397,2,1,2,0,0
400 | 398,0,7,0,11,0
401 | 399,6,13,30,70,0
402 | 400,3,1,7,0,0
403 | 401,8,8,33,32,16
404 | 402,6,11,300,71,0
405 | 403,1,17,21,195,2
406 | 404,2,0,12,1,6
407 | 405,1,2,4,10,0
408 | 406,10,12,141,47,3
409 | 407,5,2,4,0,0
410 | 408,9,5,45,2,34
411 | 409,4,0,15,7,1
412 | 410,1,3,0,0,0
413 | 411,17,0,329,345,0
414 | 412,5,13,105,61,0
415 | 413,8,2,0,0,0
416 | 414,3,7,10,13,0
417 | 415,4,0,20,2,0
418 | 416,2,34,7,341,0
419 | 417,6,2,4,0,0
420 | 418,11,6,139,150,0
421 | 419,3,1,0,0,0
422 | 420,8,6,13,4,16
423 | 421,1,3,0,0,0
424 | 422,10,0,93,19,90
425 | 423,0,4,0,4,0
426 | 424,1,0,34,18,0
427 | 425,0,1,14,16,0
428 | 426,2,5,6,32,5
429 | 427,1,3,4,28,0
430 | 428,1,0,1,1,0
431 | 429,3,18,19,25,0
432 | 430,4,0,10,6,1
433 | 431,0,2,5,8,0
434 | 432,4,8,23,23,0
435 | 433,0,3,5,41,0
436 | 434,7,3,21,101,0
437 | 435,0,4,1,19,0
438 | 436,2,1,0,0,0
439 | 437,36,1,215,4,192
440 | 438,15,6,169,0,123
441 | 439,2,3,21,16,0
442 | 440,2,13,4,7,0
443 | 441,1,0,8,16,0
444 | 442,1,2,5,23,0
445 | 443,4,1,12,35,0
446 | 444,1,4,0,0,0
447 | 445,1,0,11,16,0
448 | 446,9,1,62,19,32
449 | 447,0,4,0,2,0
450 | 448,1,1,2,1,1
451 | 449,1,6,0,0,0
452 | 450,1,0,0,0,0
453 | 451,2,3,44,29,0
454 | 452,35,7,207,9,100
455 | 453,2,1,9,20,0
456 | 454,2,5,22,35,1
457 | 455,3,0,49,17,2
458 | 456,3,3,5,0,0
459 | 457,2,1,5,13,0
460 | 458,1,9,7,6,0
461 | 459,15,7,58,3,30
462 | 460,1,1,22,21,0
463 | 461,1,2,17,40,0
464 | 462,15,6,59,6,59
465 | 463,6,0,57,7,7
466 | 464,0,1,0,16,0
467 | 465,10,2,94,172,0
468 | 466,3,1,9,3,7
469 | 467,1,4,0,2,0
470 | 468,8,14,39,10,0
471 | 469,0,2,8,7,0
472 | 470,0,0,1,1,0
473 | 471,4,2,7,4,0
474 | 472,6,4,45,15,0
475 | 473,2,0,19,57,0
476 | 474,1,2,7,26,0
477 | 475,3,6,61,85,0
478 | 476,9,0,46,39,0
479 | 477,9,5,50,16,14
480 | 478,8,9,49,13,44
481 | 479,5,5,0,0,0
482 | 480,3,2,30,9,1
483 | 481,3,2,4,45,0
484 | 482,7,1,44,7,33
485 | 483,1,3,3,31,0
486 | 484,6,40,0,5,0
487 | 485,10,6,45,35,0
488 | 486,3,2,0,1,1
489 | 487,3,1,0,0,0
490 | 488,6,11,145,78,0
491 | 489,2,0,2,16,0
492 | 490,6,5,14,4,0
493 | 491,0,1,9,4,0
494 | 492,2,2,9,21,1
495 | 493,0,21,0,65,0
496 | 494,24,44,72,19,53
497 | 495,5,0,20,2,20
498 | 496,0,2,5,0,4
499 | 497,1,2,0,0,0
500 | 498,0,1,1,19,0
501 | 499,3,3,2,0,0
502 | 500,4,5,0,2,0
503 | 501,2,0,2,3,0
504 | 502,6,3,59,32,0
505 | 503,0,5,6,31,0
506 | 504,2,0,0,3,0
507 | 505,1,9,0,42,0
508 | 506,0,0,9,21,0
509 | 507,8,6,27,47,0
510 | 508,0,0,4,1,0
511 | 509,23,11,87,63,44
512 | 510,2,1,0,0,0
513 | 511,3,4,16,27,0
514 | 512,1,4,0,0,0
515 | 513,7,0,8,2,7
516 | 514,6,3,260,109,0
517 | 515,0,8,2,1,0
518 | 516,17,6,93,14,61
519 | 517,2,1,4,0,0
520 | 518,4,7,59,31,0
521 | 519,6,3,44,17,13
522 | 520,6,2,21,0,0
523 | 521,1,0,7,5,0
524 | 522,5,3,67,43,5
525 | 523,16,4,64,13,16
526 | 524,5,4,1,1,0
527 | 525,3,0,9,35,0
528 | 526,4,10,20,19,13
529 | 527,2,0,29,6,1
530 | 528,8,0,30,4,24
531 | 529,5,3,16,31,0
532 | 530,3,1,24,16,0
533 | 531,0,9,5,6,0
534 | 532,2,0,0,0,0
535 | 533,2,2,0,2,0
536 | 534,0,1,0,0,0
537 | 535,7,0,81,36,7
538 | 536,15,1,52,4,44
539 | 537,1,4,2,8,0
540 | 538,10,3,162,9,115
541 | 539,2,5,0,0,0
542 | 540,22,21,40,11,1
543 | 541,4,1,12,2,0
544 | 542,3,2,0,0,0
545 | 543,4,2,0,6,0
546 | 544,4,5,8,2,0
547 | 545,3,14,36,21,0
548 | 546,3,1,16,47,0
549 | 547,2,2,18,34,0
550 | 548,3,0,5,4,0
551 | 549,1,4,31,0,0
552 | 550,2,0,11,14,0
553 | 551,1,0,13,13,0
554 | 552,2,3,52,45,0
555 | 553,6,3,3,1,0
556 | 554,2,9,9,1,1
557 | 555,5,3,20,28,0
558 | 556,13,17,133,32,102
559 | 557,0,3,0,3,0
560 | 558,14,3,87,21,39
561 | 559,2,2,29,41,0
562 | 560,1,2,0,38,0
563 | 561,3,0,33,42,0
564 | 562,3,4,31,68,0
565 | 563,6,5,68,6,88
566 | 564,1,0,0,0,0
567 | 565,0,0,6,8,0
568 | 566,2,2,0,0,0
569 | 567,5,3,66,30,1
570 | 568,4,0,9,19,2
571 | 569,19,9,113,24,105
572 | 570,2,3,0,0,0
573 | 571,3,0,30,13,0
574 | 572,0,0,11,9,0
575 | 573,2,5,0,0,0
576 | 574,4,4,22,9,7
577 | 575,2,1,19,3,0
578 | 576,3,12,24,24,0
579 | 577,3,1,117,97,0
580 | 578,6,3,26,0,16
581 | 579,0,0,24,33,0
582 | 580,3,5,0,0,0
583 | 581,21,10,130,15,16
584 | 582,1,4,0,0,0
585 | 583,2,2,3,1,0
586 | 584,7,0,4,0,3
587 | 585,2,4,5,20,0
588 | 586,3,0,1,5,1
589 | 587,31,14,200,77,274
590 | 588,5,31,13,11,0
591 | 589,1,1,21,30,0
592 | 590,47,16,322,41,300
593 | 591,22,14,134,116,43
594 | 592,1,1,0,0,0
595 | 593,1,2,32,5,0
596 | 594,0,1,1,0,0
597 | 595,2,0,30,15,0
598 | 596,5,4,30,47,0
599 | 597,0,0,1,0,0
600 | 598,4,5,32,15,0
601 | 599,3,8,1,5,0
602 | 600,0,1,19,27,0
603 | 601,9,4,38,92,0
604 | 602,3,17,0,0,0
605 | 603,2,3,1,0,0
606 | 604,12,5,105,204,2
607 | 605,4,6,2,5,0
608 | 606,16,5,288,11,211
609 | 607,2,3,15,4,0
610 | 608,2,0,12,24,0
611 | 609,1,5,0,0,0
612 | 610,5,5,17,12,0
613 | 611,2,1,32,13,0
614 | 612,3,0,1,0,0
615 | 613,4,6,15,2,1
616 | 614,1,0,5,10,0
617 | 615,2,1,35,22,0
618 | 616,8,18,79,25,0
619 | 617,14,12,88,47,38
620 | 618,5,23,98,25,0
621 | 619,6,2,27,6,12
622 | 620,5,6,20,1,21
623 | 621,15,12,53,36,30
624 | 622,6,2,32,1,38
625 | 623,6,0,0,0,0
626 | 624,7,6,38,1,20
627 | 625,28,5,149,11,129
628 | 626,1,0,0,0,0
629 | 627,1,1,0,0,0
630 | 628,11,4,18,1,0
631 | 629,13,12,75,1,57
632 | 630,2,0,0,0,0
633 | 631,2,2,0,0,0
634 | 632,5,3,1,6,0
635 | 633,5,1,39,15,25
636 | 634,1,0,0,0,0
637 | 635,3,1,20,7,0
638 | 636,2,2,34,2,0
639 | 637,6,18,85,17,40
640 | 638,2,2,0,0,0
641 | 639,4,12,31,81,2
642 | 640,8,3,33,8,0
643 | 641,1,2,70,102,0
644 | 642,1,9,0,0,0
645 | 643,1,6,0,29,0
646 | 644,2,2,22,9,0
647 | 645,10,7,73,51,67
648 | 646,1,1,11,3,0
649 | 647,1,2,0,1,0
650 | 648,5,0,56,0,41
651 | 649,2,0,19,49,0
652 | 650,2,25,0,13,0
653 | 651,1,7,11,26,0
654 | 652,0,3,0,0,0
655 | 653,2,3,0,18,0
656 | 654,1,3,0,8,0
657 | 655,5,3,22,1,17
658 | 656,4,6,14,7,9
659 | 657,0,1,0,0,0
660 | 658,5,8,0,0,0
661 | 659,1,1,21,25,0
662 | 660,1,0,1,4,0
663 | 661,2,3,17,49,0
664 | 662,10,6,245,20,0
665 | 663,1,4,9,11,0
666 | 664,18,15,129,67,71
667 | 665,0,9,0,2,0
668 | 666,0,4,4,14,0
669 | 667,4,1,1,0,1
670 | 668,0,3,4,32,0
671 | 669,5,4,16,0,0
672 | 670,0,4,21,19,0
673 | 671,1,1,11,58,0
674 | 672,5,25,0,10,0
675 | 673,1,0,0,16,0
676 | 674,20,3,112,23,118
677 | 675,4,1,5,0,5
678 | 676,16,10,43,6,22
679 | 677,7,6,6,4,2
680 | 678,3,7,19,45,0
681 | 679,15,3,94,16,83
682 | 680,17,5,160,8,162
683 | 681,2,1,24,9,0
684 | 682,1,2,6,9,0
685 | 683,1,2,3,2,0
686 | 684,17,4,88,45,54
687 | 685,6,12,0,0,0
688 | 686,3,36,0,59,0
689 | 687,6,4,72,9,3
690 | 688,0,8,0,13,0
691 | 689,1,0,1,0,0
692 | 690,5,5,34,7,28
693 | 691,1,1,0,0,0
694 | 692,0,0,15,45,0
695 | 693,3,10,7,2,0
696 | 694,3,1,5,1,0
697 | 695,1,0,2,0,2
698 | 696,2,4,0,2,0
699 | 697,1,1,2,0,0
700 | 698,1,1,0,14,0
701 | 699,20,13,39,2,14
702 | 700,9,4,82,9,23
703 | 701,12,10,91,19,158
704 | 702,8,4,25,10,10
705 | 703,7,0,38,52,0
706 | 704,26,6,168,25,105
707 | 705,3,0,19,38,2
708 | 706,2,5,39,18,0
709 | 707,4,22,1,7,0
710 | 708,3,7,0,3,0
711 | 709,1,2,19,50,0
712 | 710,7,3,0,0,0
713 | 711,4,7,5,3,0
714 | 712,3,0,52,64,0
715 | 713,2,3,10,5,0
716 | 714,13,2,85,7,21
717 | 715,9,1,57,7,35
718 | 716,5,5,53,30,0
719 | 717,1,4,16,2,1
720 | 718,13,3,39,1,42
721 | 719,1,3,15,10,0
722 | 720,8,1,51,7,24
723 | 721,6,1,47,13,0
724 | 722,19,6,164,7,109
725 | 723,8,0,41,5,45
726 | 724,6,40,4,87,0
727 | 725,6,2,68,100,0
728 | 726,3,4,3,19,0
729 | 727,13,6,103,18,66
730 | 728,13,4,104,11,77
731 | 729,1,4,3,29,0
732 | 730,1,3,0,0,0
733 | 731,4,0,12,0,8
734 | 732,4,8,0,0,0
735 | 733,1,0,0,0,0
736 | 734,1,1,0,0,0
737 | 735,0,3,9,0,0
738 | 736,9,36,0,2,0
739 | 737,2,3,0,0,0
740 | 738,1,1,0,8,0
741 | 739,4,6,1,0,1
742 | 740,9,4,20,8,2
743 | 741,32,5,228,38,176
744 | 742,6,1,23,6,14
745 | 743,12,5,113,13,91
746 | 744,6,3,51,498,0
747 | 745,32,10,270,30,183
748 | 746,1,0,11,19,0
749 | 747,17,6,251,17,225
750 | 748,0,6,0,0,0
751 | 749,6,0,24,8,2
752 | 750,5,1,9,7,0
753 | 751,0,7,0,2,0
754 | 752,6,1,16,6,12
755 | 753,8,1,35,19,18
756 | 754,2,0,19,31,0
757 | 755,5,17,51,139,0
758 | 756,0,0,20,34,0
759 | 757,1,0,0,0,0
760 | 758,3,0,12,85,0
761 | 759,19,5,143,13,119
762 | 760,3,15,0,7,0
763 | 761,9,19,53,61,0
764 | 762,2,2,2,0,0
765 | 763,0,1,0,74,0
766 | 764,4,4,26,2,7
767 | 765,0,1,20,19,0
768 | 766,1,1,1,0,0
769 | 767,6,7,53,14,0
770 | 768,2,1,0,0,0
771 | 769,12,0,161,15,130
772 | 770,1,4,24,5,0
773 | 771,7,1,113,28,0
774 | 772,0,1,0,1,0
775 | 773,28,8,213,5,138
776 | 774,20,0,139,12,114
777 | 775,3,4,51,4,49
778 | 776,3,2,25,2,29
779 | 777,2,3,0,0,0
780 | 778,2,3,0,1,0
781 | 779,1,0,0,0,0
782 | 780,0,0,3,3,0
783 | 781,3,1,43,7,0
784 | 782,0,3,95,22,0
785 | 783,17,8,88,18,10
786 | 784,11,5,211,211,0
787 | 785,41,8,288,22,230
788 | 786,1,2,0,0,0
789 | 787,3,4,0,10,0
790 | 788,4,19,12,5,0
791 | 789,3,2,23,1,18
792 | 790,7,2,41,4,36
793 | 791,4,4,37,3,26
794 | 792,10,0,49,1,42
795 | 793,11,5,8,0,0
796 | 794,4,3,89,106,0
797 | 795,3,0,8,0,1
798 | 796,1,0,1,8,0
799 | 797,1,0,0,0,0
800 | 798,2,1,0,1,0
801 | 799,4,12,13,74,3
802 | 800,1,1,0,0,0
803 | 801,9,3,56,30,57
804 | 802,11,6,48,54,12
805 | 803,21,10,258,40,151
806 | 804,6,1,56,52,0
807 | 805,12,1,60,14,32
808 | 806,5,12,14,4,10
809 | 807,7,0,57,7,56
810 | 808,1,0,0,0,0
811 | 809,3,0,48,43,1
812 | 810,15,4,68,28,20
813 | 811,7,5,12,0,0
814 | 812,3,0,16,3,0
815 | 813,1,3,6,21,0
816 | 814,1,6,0,6,0
817 | 815,5,3,2,4,0
818 | 816,3,0,35,33,0
819 | 817,4,5,26,39,0
820 | 818,1,0,4,7,0
821 | 819,2,0,0,0,0
822 | 820,4,3,44,4,41
823 | 821,1,0,3,2,0
824 | 822,30,10,305,12,210
825 | 823,2,10,29,12,0
826 | 824,2,4,0,0,0
827 | 825,3,1,0,1,0
828 | 826,0,2,3,0,0
829 | 827,8,7,4,0,0
830 | 828,15,9,49,4,35
831 | 829,4,3,15,9,8
832 | 830,2,10,4,32,0
833 | 831,3,2,6,1,0
834 | 832,0,4,0,18,0
835 | 833,22,3,226,43,215
836 | 834,0,7,0,1,0
837 | 835,4,6,1,2,0
838 | 836,2,3,3,1,0
839 | 837,1,0,3,4,0
840 | 838,11,28,2,5,0
841 | 839,4,1,1,1,0
842 | 840,2,3,47,14,0
843 | 841,3,6,0,26,0
844 | 842,2,4,6,4,0
845 | 843,3,0,6,4,0
846 | 844,9,7,5,4,0
847 | 845,8,1,9,16,0
848 | 846,1,1,13,3,0
849 | 847,27,3,227,46,151
850 | 848,1,1,5,23,0
851 | 849,6,1,24,5,10
852 | 850,6,13,13,0,0
853 | 851,1,2,10,8,0
854 | 852,6,27,3,10,3
855 | 853,0,4,0,0,0
856 | 854,1,2,0,1,0
857 | 855,3,2,24,78,0
858 | 856,8,4,14,5,0
859 | 857,1,10,0,3,0
860 | 858,6,12,20,67,0
861 | 859,3,6,8,33,0
862 | 860,3,2,20,11,7
863 | 861,4,6,110,0,0
864 | 862,4,2,18,99,1
865 | 863,1,1,0,0,0
866 | 864,14,2,207,20,172
867 | 865,1,0,0,1,1
868 | 866,2,0,4,14,0
869 | 867,0,2,5,11,0
870 | 868,5,11,62,94,0
871 | 869,2,3,0,0,0
872 | 870,2,0,7,0,0
873 | 871,23,10,196,43,176
874 | 872,7,11,132,36,0
875 | 873,1,0,30,77,0
876 | 874,18,7,118,47,117
877 | 875,4,8,60,14,0
878 | 876,1,0,0,0,0
879 | 877,1,3,7,3,0
880 | 878,1,1,16,36,0
881 | 879,0,0,28,75,0
882 | 880,2,0,9,23,0
883 | 881,43,16,169,59,117
884 | 882,2,4,6,16,0
885 | 883,1,3,0,0,0
886 | 884,3,0,59,13,7
887 | 885,3,0,21,12,0
888 | 886,2,0,30,0,15
889 | 887,1,2,0,0,0
890 | 888,1,0,5,4,0
891 | 889,11,1,32,4,15
892 | 890,3,0,10,3,0
893 | 891,7,0,27,5,22
894 | 892,16,6,144,3,124
895 | 893,1,3,30,35,8
896 | 894,2,2,41,39,0
897 | 895,3,2,52,15,7
898 | 896,4,0,0,0,0
899 | 897,0,4,21,11,0
900 | 898,10,4,79,6,45
901 | 899,0,3,0,2,0
902 | 900,42,5,682,26,529
903 | 901,3,11,123,31,0
904 | 902,0,2,0,0,0
905 | 903,4,5,30,14,1
906 | 904,1,1,0,0,0
907 | 905,3,2,6,0,0
908 | 906,4,2,20,7,2
909 | 907,3,3,32,55,0
910 | 908,4,2,0,0,0
911 | 909,5,4,30,9,0
912 | 910,3,16,0,5,0
913 | 911,4,5,149,50,0
914 | 912,30,2,244,13,205
915 | 913,1,0,20,33,0
916 | 914,3,23,15,115,0
917 | 915,0,0,2,5,0
918 | 916,1,3,12,13,0
919 | 917,10,8,32,110,0
920 | 918,4,2,25,7,0
921 | 919,5,3,41,64,3
922 | 920,11,2,280,212,2
923 | 921,1,0,25,51,0
924 | 922,4,4,7,6,0
925 | 923,8,2,106,11,86
926 | 924,23,3,131,8,125
927 | 925,3,3,18,62,0
928 | 926,2,0,3,32,0
929 | 927,1,0,17,6,0
930 | 928,1,0,0,22,0
931 | 929,3,0,23,50,0
932 | 930,8,0,36,8,21
933 | 931,3,2,20,17,0
934 | 932,5,8,21,67,0
935 | 933,6,7,68,70,21
936 | 934,4,2,36,17,0
937 | 935,0,1,0,0,0
938 | 936,1,0,1,2,0
939 | 937,4,10,26,10,0
940 | 938,0,0,4,0,0
941 | 939,16,3,98,8,66
942 | 940,2,4,3,74,0
943 | 941,6,2,47,2,39
944 | 942,1,6,0,37,0
945 | 943,1,1,24,26,0
946 | 944,1,4,1,0,0
947 | 945,6,0,79,6,45
948 | 946,3,1,34,33,0
949 | 947,6,4,29,6,24
950 |
--------------------------------------------------------------------------------
/dataset.py:
--------------------------------------------------------------------------------
1 | """ Kaggle Sealion Pytorch Dataset
2 | Pytorch Dataset code for patched based training and prediction of the
3 | NOAA Fishes Sea Lion counting Kaggle data.
4 |
5 | Dataset code generates or loads targets for density and counception
6 | based counting models.
7 | """
8 | from collections import defaultdict
9 | import cv2
10 | import torch
11 | import torch.utils.data as data
12 | from torch.utils.data.sampler import Sampler
13 | from torchvision import datasets, transforms
14 | from PIL import Image
15 | import random
16 | import pandas as pd
17 | import numpy as np
18 | import os
19 | import functools
20 | import time
21 | import mytransforms
22 | import utils
23 |
24 | IMG_EXTENSIONS = ['.jpg', '.jpeg', '.png']
25 | CATEGORIES = ["adult_males", "subadult_males", "adult_females", "juveniles", "pups"]
26 | CATEGORY_MAP = {"adult_males": 0, "subadult_males": 1, "adult_females": 2, "juveniles": 3, "pups": 4}
27 | TARGET_TYPES = ['density', 'countception']
28 |
29 |
30 | def to_tensor(arr):
31 | assert(isinstance(arr, np.ndarray))
32 | t = torch.from_numpy(arr.transpose((2, 0, 1)))
33 | if isinstance(t, torch.ByteTensor):
34 | return t.float().div(255)
35 | return t
36 |
37 |
38 | def find_inputs(folder, types=IMG_EXTENSIONS, extract_extra=False):
39 | inputs = []
40 | for root, _, files in os.walk(folder, topdown=False):
41 | for rel_filename in files:
42 | base, ext = os.path.splitext(rel_filename)
43 | if ext.lower() in types:
44 | abs_filename = os.path.join(root, rel_filename)
45 | if extract_extra:
46 | img = Image.open(abs_filename)
47 | if not img:
48 | continue
49 | w, h = img.size
50 | info = dict(filename=abs_filename, width=w, height=h, xmin=0, ymin=0, xmax=w, ymax=h)
51 | else:
52 | info = dict(filename=abs_filename)
53 | inputs.append((int(base), info))
54 | if inputs:
55 | return zip(*sorted(inputs, key=lambda k: k[0]))
56 | else:
57 | return [], []
58 |
59 |
60 | def find_targets(folder, input_ids, types=IMG_EXTENSIONS):
61 | inputs_set = set(input_ids)
62 | targets = defaultdict(dict)
63 | for root, _, files in os.walk(folder, topdown=False):
64 | for rel_filename in files:
65 | base, ext = os.path.splitext(rel_filename)
66 | if ext.lower() in types:
67 | split = base.split('-')
68 | fid = int(split[0])
69 | if fid in inputs_set:
70 | abs_filename = os.path.join(root, rel_filename)
71 | if len(split) > 2:
72 | targets[fid][int(split[2])] = abs_filename
73 | else:
74 | targets[fid] = abs_filename
75 | return targets
76 |
77 |
78 | def gen_target_gauss(coords, size, sigma=5, kernel_size=(21, 21), factor=1024.):
79 | w, h = size
80 | num_outputs = len(CATEGORIES)
81 | target_img = np.zeros(shape=(h, w, num_outputs), dtype=np.float32)
82 | for cat_idx, cat_name in enumerate(CATEGORIES):
83 | xy = coords[coords[:, 2] == cat_idx][:, :2]
84 | for x, y in xy:
85 | target_img[y, x, cat_idx] += factor
86 | target_img = cv2.GaussianBlur(target_img, kernel_size, sigma, borderType=cv2.BORDER_CONSTANT)
87 | return target_img
88 |
89 |
90 | def gen_target_countception(coords, size, subpatch_size=32, stride=1, max_count=0, dtype=np.float32):
91 | w, h = size
92 | pad = (subpatch_size - 1) // 2
93 | w = (w + 2 * pad) // stride
94 | h = (h + 2 * pad) // stride
95 | #print(size, w, h)
96 | num_outputs = len(CATEGORIES)
97 | coords_pad = coords.copy()
98 | coords_pad[:, :2] = coords[:, :2] + [subpatch_size, subpatch_size]
99 | target_img = np.zeros(shape=(h, w, num_outputs), dtype=dtype)
100 | for x in range(w):
101 | for y in range(h):
102 | subpatch_points = utils.crop_points(coords, x * stride, y * stride, subpatch_size, subpatch_size)
103 | for p in subpatch_points:
104 | target_img[y][x][p[2]] += dtype(1)
105 | #print(target_img.sum(axis=(0, 1))/(subpatch_size**2), target_img.max())
106 | if max_count > 0:
107 | target_img = np.clip(target_img, 0, max_count)
108 | return target_img
109 |
110 |
111 | def gen_mask(input_img, dotted_file):
112 | img_dotted = cv2.imread(dotted_file)
113 | mask = cv2.cvtColor(img_dotted, cv2.COLOR_BGR2GRAY)
114 | _, mask = cv2.threshold(mask, 10, 255, cv2.THRESH_BINARY)
115 | img_masked = cv2.bitwise_and(input_img, input_img, mask=mask)
116 | return img_masked, mask
117 |
118 |
119 | class ImagePatchIndex:
120 | def __init__(self, image_index, patch_index=0):
121 | self.image_index = image_index
122 | self.patch_index = patch_index
123 |
124 |
125 | class IndexedPatchSampler(Sampler):
126 | """Samples patches across images sequentially by index in raster order.
127 | """
128 |
129 | def __init__(self, data_source):
130 | self.num_images = len(data_source)
131 | if data_source.patch_count:
132 | self.num_patches = data_source.patch_count
133 | self.patch_index = data_source.patch_index
134 | else:
135 | # fallback to indexing whole images from dataset
136 | print('Warning: Data source has no patch information, falling back to whole image indexing.')
137 | self.num_patches = 0
138 | self.patch_index = []
139 |
140 | def __iter__(self):
141 | if self.num_patches:
142 | for i in range(self.num_images):
143 | for j in self.patch_index[i]:
144 | yield ImagePatchIndex(i, j)
145 | else:
146 | for i in range(self.num_images):
147 | yield i
148 |
149 | def __len__(self):
150 | return self.num_patches if self.num_patches else self.num_images
151 |
152 |
153 | class RandomPatchSampler(Sampler):
154 | """Oversamples random patches from images in random order.
155 | Repeats the same image index multiple times in a row to sample 'repeat' times
156 | from the same image for big read efficiency gains.
157 | """
158 | def __init__(self, data_source, oversample=32, repeat=1):
159 | self.oversample = oversample//repeat * repeat
160 | self.repeat = repeat
161 | self.num_samples = len(data_source)
162 |
163 | def __iter__(self):
164 | # There are simpler/more compact ways of doing this, but why not have a somewhat
165 | # meaningful fake patch index?
166 | for to in range(self.oversample//self.repeat):
167 | samples = torch.randperm(self.num_samples).long()
168 | for image_index in samples:
169 | for ti in range(self.repeat):
170 | fake_patch_index = to * self.repeat + ti
171 | yield ImagePatchIndex(image_index, fake_patch_index)
172 |
173 | def __len__(self):
174 | return self.num_samples * self.oversample
175 |
176 |
177 | class SealionDataset(data.Dataset):
178 | def __init__(
179 | self,
180 | input_root,
181 | target_root='',
182 | counts_file='',
183 | coords_file='',
184 | processing_file='',
185 | train=True,
186 | patch_size=(256, 256),
187 | patch_stride=128,
188 | prescale=0.0,
189 | generate_target=True,
190 | target_type='density',
191 | per_image_norm=False,
192 | num_logits=0,
193 | transform=None,
194 | target_transform=None):
195 |
196 | extract_extra = False if os.path.exists(processing_file) else True
197 | input_ids, input_infos = find_inputs(
198 | input_root, types=['.jpg'], extract_extra=extract_extra)
199 | if len(input_ids) == 0:
200 | raise(RuntimeError("Found 0 images in : " + input_root))
201 | self.input_index = input_ids
202 |
203 | self.patch_index = [[]] * len(input_ids)
204 | self.patch_count = 0
205 | self.patch_size = patch_size
206 | self.patch_stride = patch_stride
207 | self.prescale = prescale if prescale != 1.0 else 0.0
208 | assert target_type in TARGET_TYPES
209 | self.target_type = target_type
210 | self.num_logits = num_logits
211 | if train and num_logits:
212 | assert target_type == 'countception'
213 | self.generate_target = generate_target # generate on the fly instead of loading
214 |
215 | self.data_by_id = dict()
216 | for index, (k, v) in enumerate(zip(input_ids, input_infos)):
217 | if 'width' in v:
218 | if self.prescale:
219 | v = self._apply_prescale(v, self.prescale)
220 | patch_info = self._calc_patch_info(v)
221 | num_patches = patch_info['num']
222 | self.patch_index[index] = list(range(num_patches))
223 | self.patch_count += num_patches
224 | v['patches'] = patch_info
225 | v['index'] = index
226 | self.data_by_id[k] = v
227 |
228 | self.has_targets = False
229 | if os.path.exists(target_root):
230 | targets = find_targets(target_root, input_ids, types=['.npz'])
231 | if len(targets):
232 | for k, v in targets.items():
233 | self.data_by_id[k]['target'] = v
234 | self.has_targets = True
235 | else:
236 | raise (RuntimeError("Found 0 targets in : " + target_root))
237 |
238 | if train:
239 | assert self.has_targets
240 | self.train = train
241 |
242 | if counts_file:
243 | counts_df = pd.read_csv(counts_file).rename(columns=CATEGORY_MAP)
244 | counts_df.drop(['train_id'], 1, inplace=True)
245 | for k, v in counts_df.to_dict(orient='index').items():
246 | if k in self.data_by_id:
247 | d = self.data_by_id[k]
248 | d['counts_by_cat'] = v
249 | d['count'] = sum(v.values())
250 |
251 | if processing_file:
252 | process_df = pd.read_csv(processing_file, index_col=False)
253 | cols = ['xmin', 'ymin', 'xmax', 'ymax', 'width', 'height']
254 | process_df[cols] = process_df[cols].astype(int)
255 | process_df['train_id'] = process_df.filename.map(lambda x: int(os.path.splitext(x)[0]))
256 | process_df.set_index(['train_id'], inplace=True)
257 | for k, v in process_df[cols].to_dict(orient='index').items():
258 | if k in self.data_by_id:
259 | d = self.data_by_id[k]
260 | if self.prescale:
261 | v = self._apply_prescale(v, self.prescale)
262 | patch_info = self._calc_patch_info(v)
263 | num_patches = patch_info['num']
264 | self.patch_index[d['index']] = list(range(num_patches))
265 | self.patch_count += num_patches
266 | v['patches'] = patch_info
267 | d.update(v)
268 | #print(d, self.patch_count)
269 |
270 | if coords_file:
271 | coords_df = pd.read_csv(coords_file, index_col=False)
272 | coords_df.x_coord = coords_df.x_coord.astype('int')
273 | coords_df.y_coord = coords_df.y_coord.astype('int')
274 | coords_df.category = coords_df.category.replace(CATEGORY_MAP)
275 | groupby_file = coords_df.groupby(['filename'])
276 | for file in groupby_file.indices:
277 | coords = groupby_file.get_group(file)
278 | coords = coords[['x_coord', 'y_coord', 'category']].as_matrix()
279 | coords = coords[coords[:, 0].argsort()]
280 | fid = int(os.path.splitext(file)[0])
281 | if fid in self.data_by_id:
282 | d = self.data_by_id[fid]
283 | if self.prescale:
284 | coords[:, :2] = np.rint(coords[:, :2] * self.prescale)
285 | xy_offset = np.array([d['xmin'], d['ymin']])
286 | coords[:, :2] = coords[:, :2] + xy_offset
287 | d['coords'] = coords
288 |
289 | self.dataset_mean = [0.43632373, 0.46022959, 0.4618598]
290 | self.dataset_std = [0.17749958, 0.16631233, 0.16272708]
291 | if transform is None:
292 | tfs = []
293 | if per_image_norm:
294 | tfs.append(mytransforms.NormalizeImg())
295 | tfs.append(mytransforms.ToTensor())
296 | if self.train:
297 | tfs.append(mytransforms.ColorJitter())
298 | if not per_image_norm:
299 | tfs.append(transforms.Normalize(self.dataset_mean, self.dataset_std))
300 | self.transform = transforms.Compose(tfs)
301 | self.target_transform = target_transform
302 | self.ttime = utils.AverageMeter()
303 |
304 | def _apply_prescale(self, input_info, scale):
305 | for k in ['xmin', 'xmax', 'ymin', 'ymax', 'width', 'height']:
306 | input_info[k] = np.rint(input_info[k] * scale).astype(np.int)
307 | return input_info
308 |
309 | def _calc_patch_info(self, input_info):
310 | x_min = input_info['xmin']
311 | x_max = input_info['xmax']
312 | y_min = input_info['ymin']
313 | y_max = input_info['ymax']
314 | assert y_max > y_min and x_max > x_min
315 | buffer_w = input_info['width']
316 | buffer_h = input_info['height']
317 | box_w = x_max - x_min
318 | box_h = y_max - y_min
319 | # FIXME switch to use bbox constraints
320 | num_patches, patch_cols, patch_rows = utils.calc_num_patches(
321 | buffer_w, buffer_h, self.patch_size, self.patch_stride)
322 | patch_origin_x = 0
323 | patch_origin_y = 0
324 | # if we have a bounding box border, see if we can squeeze an extra box in each dimension
325 | # if x_min != 0 or x_max != buffer_w:
326 | # new_w = patch_cols * stride + patch_size[0]
327 | # print(new_w, buffer_w)
328 | # if new_w <= buffer_w:
329 | # patch_cols += 1
330 | # patch_origin_x = x_min - (new_w - box_w) // 2
331 | # if y_min != 0 or y_max != buffer_h:
332 | # new_h = patch_rows * stride + patch_size[1]
333 | # if new_h <= buffer_h:
334 | # patch_rows += 1
335 | # patch_origin_y = y_min - (new_h - box_h) // 2
336 | num_patches = patch_cols * patch_rows
337 | patch_info = dict(
338 | num=num_patches, cols=patch_cols, rows=patch_rows, origin_x=patch_origin_x, origin_y=patch_origin_y)
339 | return patch_info
340 |
341 | @functools.lru_cache(4)
342 | def _load_input(self, input_id):
343 | path = self.data_by_id[input_id]['filename']
344 | print("Loading %s" % path)
345 | img = cv2.imread(path)
346 | if self.prescale:
347 | dsize = (self.data_by_id[input_id]['width'], self.data_by_id[input_id]['height'])
348 | img = cv2.resize(img, dsize)
349 | img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
350 | #h, w = img.shape[:2]
351 | #bh = ((h - 1) // self.patch_size[1] + 1) * self.patch_size[1] - h
352 | #bw = ((w - 1) // self.patch_size[0] + 1) * self.patch_size[0] - w
353 | #if bh or bw:
354 | # bwl = bw // 2
355 | # bhl = bh // 2
356 | # print("Adding border...", bh, bw)
357 | # img = cv2.copyMakeBorder(img, bhl, bh-bhl, bwl, bw-bwl, cv2.BORDER_CONSTANT, (0, 0, 0))
358 | # print('%d -> %d x %d -> %d' % (w, img.shape[1], h, img.shape[0]))
359 | return img
360 |
361 | @functools.lru_cache(4)
362 | def _load_target(self, input_id):
363 | d = self.data_by_id[input_id]
364 | if isinstance(d['target'], dict):
365 | tp = [cv2.imread(d['target'][x], -1) for x in range(5)]
366 | target = np.dstack(tp)
367 | target = target / np.iinfo(np.uint16).max
368 | target = target.astype(np.float32, copy=False)
369 | else:
370 | target = np.load(d['target'])['arr_0']
371 | target = target.astype(np.float32, copy=False)
372 | return target
373 |
374 | def _indexed_patch_center(self, input_id, patch_index):
375 | d = self.data_by_id[input_id]
376 | patch_info = d['patches']
377 | pc, pr = utils.index_to_rc(patch_index, patch_info['cols'])
378 | cx = pc * self.patch_stride + self.patch_size[0] // 2
379 | cy = pr * self.patch_stride + self.patch_size[1] // 2
380 | return cx, cy
381 |
382 | def _random_patch_center(self, input_id, w, h):
383 | d = self.data_by_id[input_id]
384 | if len(d['coords']) and random.random() < 0.5:
385 | # 50% of the time, randomly pick a point around an actual sealion
386 | cx, cy, _ = d['coords'][random.randint(0, len(d['coords']) - 1)]
387 | cx = cx + random.randint(-self.patch_size[0] // 4, self.patch_size[0] // 4)
388 | cy = cy + random.randint(-self.patch_size[1] // 4, self.patch_size[1] // 4)
389 | else:
390 | # return random center coords for specified patch size within a specified (x, y, w, h) bounding box
391 | pw, ph = self.patch_size[0] // 2, self.patch_size[1] // 2
392 | if 'xmin' in d:
393 | x_min = d['xmin']
394 | x_max = d['xmax']
395 | y_min = d['ymin']
396 | y_max = d['ymax']
397 | assert x_max <= w and x_max - x_min > 0
398 | assert y_max <= h and y_max - y_min > 0
399 | else:
400 | x_min = 0
401 | x_max = w
402 | y_min = 0
403 | y_max = h
404 | x_min += pw
405 | x_max -= pw
406 | y_min += ph
407 | y_max -= ph
408 | assert x_max - x_min > 0 and y_max - y_min > 0
409 | cx = random.randint(x_min, x_max)
410 | cy = random.randint(y_min, y_max)
411 | return cx, cy
412 |
413 | def _crop_and_transform(self, cx, cy, input_img, target_arr, randomize=False):
414 | target_tile = None
415 | transform_target = False if target_arr is None else True
416 | target_is_coords = True if transform_target and target_arr.shape[1] == 3 else False
417 |
418 | if randomize:
419 | angle = 0.
420 | hflip = random.random() < 0.5
421 | vflip = random.random() < 0.5
422 | do_rotate = random.random() < 0.25 if not hflip and not vflip else False
423 | if do_rotate:
424 | angle = random.random() * 360
425 | scale = random.uniform(0.5, 1.125)
426 | #print('hflip: %d, vflip: %d, angle: %f, scale: %f' % (hflip, vflip, angle, scale))
427 | else:
428 | angle = 0.
429 | scale = 1.
430 | hflip = False
431 | vflip = False
432 |
433 | crop_w, crop_h = utils.calc_crop_size(self.patch_size[0], self.patch_size[1], angle, scale)
434 | input_tile = utils.crop_center(input_img, cx, cy, crop_w, crop_h)
435 | if transform_target:
436 | if target_is_coords:
437 | target_points = target_arr.copy()
438 | target_points = utils.crop_points_center(target_points, cx, cy, crop_w, crop_h)
439 | #print(cx, cy, crop_w, crop_h, angle, scale, hflip, vflip)
440 | #print(target_points)
441 | target_points[:, :2] = target_points[:, :2] - [cx, cy]
442 | else:
443 | target_tile = utils.crop_center(target_arr, cx, cy, crop_w, crop_h)
444 |
445 | # Perform tile geometry transforms if needed
446 | if angle or scale != 1. or hflip or vflip:
447 | Mtrans = np.identity(3)
448 | Mtrans[0, 2] = (self.patch_size[0] - crop_w) // 2
449 | Mtrans[1, 2] = (self.patch_size[1] - crop_h) // 2
450 | if hflip:
451 | Mtrans[0, 0] *= -1
452 | Mtrans[0, 2] = self.patch_size[0] - Mtrans[0, 2]
453 | if vflip:
454 | Mtrans[1, 1] *= -1
455 | Mtrans[1, 2] = self.patch_size[1] - Mtrans[1, 2]
456 |
457 | if angle or scale != 1.:
458 | Mrot = cv2.getRotationMatrix2D((crop_w//2, crop_h//2), angle, scale)
459 | Mfinal = np.dot(Mtrans, np.vstack([Mrot, [0, 0, 1]]))
460 | else:
461 | Mfinal = Mtrans
462 |
463 | input_tile = cv2.warpAffine(input_tile, Mfinal[:2, :], tuple(self.patch_size))
464 | if transform_target:
465 | if target_is_coords:
466 | if len(target_points):
467 | target_cats = target_points[:, 2].copy()
468 | target_points[:, 2] = np.ones(len(target_points))
469 | target_points = np.dot(target_points, Mfinal)
470 | #print(target_points)
471 | target_points[:, 2] = target_cats
472 | else:
473 | tt64 = target_tile.astype(np.float64)
474 | tt64 = cv2.warpAffine(tt64, Mfinal[:2, :], tuple(self.patch_size))
475 | if scale != 1.:
476 | tt64 /= scale**2
477 | target_tile = tt64.astype(np.float32)
478 |
479 | if target_is_coords:
480 | target_points = np.rint(target_points).astype(np.int)
481 | target_points[:, :2] = target_points[:, :2] + [self.patch_size[0] // 2, self.patch_size[1] // 2]
482 | target_points = utils.crop_points(target_points, 0, 0, self.patch_size[0], self.patch_size[1])
483 | #print(target_points)
484 | if self.target_type == 'countception':
485 | dtype = np.uint8 if self.num_logits else np.float32
486 | max_count = self.num_logits - 1 if self.num_logits else 0
487 | target_tile = gen_target_countception(
488 | target_points, self.patch_size, max_count=max_count, dtype=dtype)
489 | else:
490 | target_tile = gen_target_gauss(target_points, self.patch_size, factor=1024.)
491 |
492 | return input_tile, target_tile
493 |
494 | def __getitem__(self, index):
495 | if isinstance(index, ImagePatchIndex):
496 | patch_index = index.patch_index
497 | index = index.image_index
498 | else:
499 | patch_index = 0 #FIXME sort this out
500 |
501 | input_id = self.input_index[index % len(self)]
502 | input_img = self._load_input(input_id)
503 | #print(input_id, index, patch_index)
504 | h, w = input_img.shape[:2]
505 | if self.train:
506 | if self.generate_target:
507 | target_arr = self.data_by_id[input_id]['coords']
508 | else:
509 | target_arr = self._load_target(input_id)
510 | #print(target_arr.shape)
511 |
512 | attempts = 2
513 | for i in range(attempts):
514 | pw, ph = self.patch_size
515 | cx, cy = self._random_patch_center(input_id, w, h)
516 | input_patch, target_patch = self._crop_and_transform(cx, cy, input_img, target_arr, randomize=True)
517 | # check centre of chosen patch_index for valid pixels
518 | if np.any(utils.crop_center(input_patch, pw//2, ph//2, pw//4, ph//4)):
519 | break
520 |
521 | input_tile_tensor = self.transform(input_patch)
522 | target_tile_tensor = to_tensor(target_patch)
523 | else:
524 | target_arr = None
525 | if self.has_targets:
526 | if self.generate_target:
527 | target_arr = self.data_by_id[input_id]['coords']
528 | else:
529 | target_arr = self._load_target(input_id)
530 |
531 | cx, cy = self._indexed_patch_center(input_id, patch_index)
532 | input_patch, target_patch = self._crop_and_transform(cx, cy, input_img, target_arr, randomize=False)
533 | input_tile_tensor = self.transform(input_patch)
534 | if target_patch is None:
535 | target_tile_tensor = torch.zeros(1)
536 | else:
537 | target_tile_tensor = to_tensor(target_patch)
538 | #print(input_tile_tensor.size(), target_tile_tensor)
539 |
540 | #cv2.imwrite('test-scaled-input-%d.png' % index, input_patch)
541 | #cv2.imwrite('test-scaled-target-%d.png' % index, 4096*target_tile[:, :, :3])
542 |
543 | index_tensor = torch.LongTensor([input_id, index, patch_index])
544 |
545 | return input_tile_tensor, target_tile_tensor, index_tensor
546 |
547 | def __len__(self):
548 | return len(self.input_index)
549 |
550 | def get_num_patches(self, input_id=None):
551 | if input_id is None:
552 | return self.patch_count
553 | else:
554 | if input_id in self.data_by_id:
555 | return self.data_by_id[input_id]['patches']['num']
556 | else:
557 | return 0
558 |
559 | def get_input_size(self, input_id):
560 | if input_id in self.data_by_id:
561 | d = self.data_by_id[input_id]
562 | return d['width'], d['height']
563 | else:
564 | return 0, 0
565 |
566 | def get_patch_cols(self, input_id):
567 | if input_id in self.data_by_id:
568 | return self.data_by_id[input_id]['patches']['cols']
569 | else:
570 | return 0
--------------------------------------------------------------------------------
/inference.py:
--------------------------------------------------------------------------------
1 | import argparse
2 | import os
3 | import time
4 | import cv2
5 | import numpy as np
6 | import pandas as pd
7 | from dataset import SealionDataset, IndexedPatchSampler
8 | from models import ModelCnet, ModelCountception
9 | from utils import AverageMeter
10 | from utils_cython import merge_patches_float32
11 | import torch
12 | import torch.autograd as autograd
13 | import torch.utils.data as data
14 |
15 |
16 | parser = argparse.ArgumentParser(description='PyTorch Sealion count inference')
17 | parser.add_argument('data', metavar='DIR',
18 | help='path to dataset')
19 | parser.add_argument('--model', default='countception', type=str, metavar='MODEL',
20 | help='Name of model to train (default: "countception"')
21 | parser.add_argument('--use-logits', action='store_true', default=False,
22 | help='Enable use of logits for model output')
23 | parser.add_argument('--patch-size', type=int, default=256, metavar='N',
24 | help='Image patch size (default: 256)')
25 | parser.add_argument('--batch-size', type=int, default=16, metavar='N',
26 | help='input batch size for training (default: 16)')
27 | parser.add_argument('--seed', type=int, default=1, metavar='S',
28 | help='random seed (default: 1)')
29 | parser.add_argument('--log-interval', type=int, default=100, metavar='N',
30 | help='how many batches to wait before logging training status')
31 | parser.add_argument('--num-processes', type=int, default=2, metavar='N',
32 | help='how many training processes to use (default: 2)')
33 | parser.add_argument('-r', '--restore-checkpoint', default=None,
34 | help='path to restore checkpoint, e.g. ./checkpoint-1.tar')
35 | parser.add_argument('--no-cuda', action='store_true', default=False,
36 | help='disables CUDA training')
37 | parser.add_argument('--num-gpu', type=int, default=1,
38 | help='Number of GPUS to use')
39 |
40 | COLS = ['test_id', 'adult_males', 'subadult_males', 'adult_females', 'juveniles', 'pups']
41 |
42 |
43 | def main():
44 | args = parser.parse_args()
45 |
46 | processed_file = os.path.join(args.data, 'processed.csv')
47 |
48 | batch_size = args.batch_size
49 | patch_size = (args.patch_size, args.patch_size)
50 | num_outputs = 5
51 | count_factor = 1024.
52 | overlapped_patches = False
53 | debug_image = False
54 | debug_model = False
55 | use_logits = args.use_logits
56 | num_logits = 12 if use_logits else 0
57 | dataset = SealionDataset(
58 | args.data,
59 | processing_file=processed_file,
60 | train=False,
61 | patch_size=patch_size,
62 | patch_stride=patch_size[0] // 2 if overlapped_patches else patch_size[0],
63 | prescale=0.5,
64 | per_image_norm=True,
65 | num_logits=num_logits)
66 | sampler = IndexedPatchSampler(dataset)
67 | loader = data.DataLoader(
68 | dataset,
69 | batch_size=batch_size,
70 | shuffle=False,
71 | num_workers=args.num_processes,
72 | sampler=sampler)
73 |
74 | if args.model == 'cnet':
75 | model = ModelCnet(
76 | outplanes=num_outputs, target_size=patch_size, debug=debug_model)
77 | elif args.model == 'countception' or args.model == 'cc':
78 | model = ModelCountception(
79 | outplanes=num_outputs, use_logits=use_logits, logits_per_output=num_logits, debug=debug_model)
80 | else:
81 | assert False and "Invalid model"
82 |
83 | if not args.no_cuda:
84 | if args.num_gpu > 1:
85 | model = torch.nn.DataParallel(model, device_ids=list(range(args.num_gpu))).cuda()
86 | else:
87 | model.cuda()
88 |
89 | if args.restore_checkpoint is not None:
90 | assert os.path.isfile(args.restore_checkpoint), '%s not found' % args.restore_checkpoint
91 | checkpoint = torch.load(args.restore_checkpoint)
92 | model.load_state_dict(checkpoint['state_dict'])
93 | print('Model restored from file: %s' % args.restore_checkpoint)
94 |
95 | model.eval()
96 |
97 | batch_time_m = AverageMeter()
98 | data_time_m = AverageMeter()
99 | current_id = -1
100 | patches = []
101 | results = []
102 | try:
103 | end = time.time()
104 | for batch_idx, (input, target, index) in enumerate(loader):
105 | data_time_m.update(time.time() - end)
106 | if not args.no_cuda:
107 | input_var, target_var = autograd.Variable(input.cuda()), autograd.Variable(target.cuda())
108 | else:
109 | input_var, target_var = autograd.Variable(input), autograd.Variable(target)
110 | output = model(input_var)
111 |
112 | if use_logits:
113 | output = torch.cat([o.max(dim=1)[1] for o in output], dim=1).float()
114 | output = output.permute(0, 2, 3, 1) / count_factor
115 | if not overlapped_patches:
116 | output = torch.squeeze(output.sum(dim=1))
117 | output = torch.squeeze(output.sum(dim=1))
118 | output = output.cpu().data.numpy()
119 |
120 | for result_index, o in zip(index, output):
121 | input_id, index, patch_index = result_index
122 | #print('input_id, index, patch_index: ', input_id, index, patch_index)
123 |
124 | if current_id == -1:
125 | current_id = input_id
126 | elif current_id != input_id:
127 | if overlapped_patches:
128 | # reconstruct output image from overlapping patches
129 | w, h = dataset.get_input_size(current_id)
130 | cols = dataset.get_patch_cols(current_id)
131 | output_arr = np.zeros((h, w, num_outputs), dtype=np.float32)
132 | patches_arr = np.stack(patches)
133 | # FIXME there are some bounds issues that need to be debuged with merge and certain image
134 | # w/h and patch/stride alignments
135 | merge_patches_float32(output_arr, patches_arr, cols, dataset.patch_size, dataset.patch_stride)
136 | counts = list(np.sum(output_arr, axis=(0, 1)))
137 | if debug_image:
138 | write_debug_img(output_arr, current_id)
139 | else:
140 | #print(len(patches))
141 | counts = list(np.sum(patches, axis=0))
142 | print(counts)
143 | results.append([current_id] + counts)
144 | patches = []
145 | current_id = input_id
146 |
147 | patches.append(o)
148 | # end iterating through batch
149 |
150 | batch_time_m.update(time.time() - end)
151 | if batch_idx % args.log_interval == 0:
152 | print('Inference: [{}/{} ({:.0f}%)] '
153 | 'Time: {batch_time.val:.3f}s, {rate:.3f}/s '
154 | '({batch_time.avg:.3f}s, {rate_avg:.3f}/s) '
155 | 'Data: {data_time.val:.3f} ({data_time.avg:.3f})'.format(
156 | batch_idx * len(input), len(loader.sampler),
157 | 100. * batch_idx / len(loader),
158 | batch_time=batch_time_m,
159 | rate=input_var.size(0) / batch_time_m.val,
160 | rate_avg=input_var.size(0) / batch_time_m.avg,
161 | data_time=data_time_m))
162 |
163 | end = time.time()
164 | #end iterating through dataset
165 | except KeyboardInterrupt:
166 | pass
167 | results_df = pd.DataFrame(results, columns=COLS)
168 | results_df.to_csv('submission.csv', index=False)
169 |
170 |
171 | def write_debug_img(img, current_id):
172 | dimg = img.astype(np.float64)
173 | dimg = (dimg[:, :, 0] + 2**8 * dimg[:, :, 1] + 2**16 * dimg[:, :, 2]
174 | + 2**24 * dimg[:, :, 3] + 2**32 * dimg[:, :, 4])
175 | dimg = cv2.normalize(
176 | dimg, None, 0, 255, norm_type=cv2.NORM_MINMAX, dtype=cv2.CV_8UC1)
177 | dimg = cv2.applyColorMap(dimg, colormap=cv2.COLORMAP_JET)
178 | cv2.imwrite('output-%d.png' % current_id, dimg)
179 |
180 |
181 | if __name__ == '__main__':
182 | main()
183 |
--------------------------------------------------------------------------------
/models/__init__.py:
--------------------------------------------------------------------------------
1 | from .model_cnet import ModelCnet
2 | from .model_countception import ModelCountception
3 |
--------------------------------------------------------------------------------
/models/model_cnet.py:
--------------------------------------------------------------------------------
1 | """ C-Net Model (Count-Net)
2 | A Pytorch model (inspired by U-net architecture) for object counting.
3 |
4 | Inspired by: https://arxiv.org/abs/1505.04597
5 | along with density counting ideas from:
6 | https://www.robots.ox.ac.uk/~vgg/publications/2015/Xie15/weidi15.pdf
7 | https://arxiv.org/pdf/1705.10118.pdf
8 | """
9 | import torch
10 | import torch.nn as nn
11 | import torch.nn.functional as F
12 | import math
13 |
14 |
15 | def conv_block(
16 | in_chan, out_chan,
17 | ksize=3, stride=1, pad=0,
18 | activation=nn.ReLU(), use_bn=False, dropout=0.):
19 | layers = []
20 | layers += [nn.Conv2d(in_chan, out_chan, kernel_size=ksize, stride=stride, padding=pad)]
21 | if use_bn:
22 | layers += [nn.BatchNorm2d(out_chan)]
23 | layers += [activation]
24 | if dropout:
25 | layers += [nn.Dropout(p=dropout)]
26 | layers += [nn.Conv2d(out_chan, out_chan, kernel_size=ksize, stride=stride, padding=pad)]
27 | if use_bn:
28 | layers += [nn.BatchNorm2d(out_chan)]
29 | layers += [activation]
30 | if dropout:
31 | layers += [nn.Dropout(p=dropout)]
32 | return nn.Sequential(*layers)
33 |
34 |
35 | def pool_layer():
36 | return nn.Sequential(nn.MaxPool2d(2))
37 |
38 |
39 | def upsample_layer(in_chan, out_chan):
40 | return nn.Sequential(
41 | nn.ConvTranspose2d(in_chan, out_chan, kernel_size=2, stride=2))
42 |
43 |
44 | class ModelCnet(nn.Module):
45 |
46 | def __init__(
47 | self,
48 | inplanes=3,
49 | outplanes=1,
50 | use_batch_norm=False,
51 | use_padding=False,
52 | target_size=(256, 256),
53 | debug=False):
54 |
55 | super(ModelCnet, self).__init__()
56 | self.inplanes = inplanes
57 | self.outplanes = outplanes
58 | self.activation = nn.LeakyReLU(0.1)
59 | self.use_batch_norm = use_batch_norm
60 | self.use_padding = use_padding
61 | self.debug = debug
62 |
63 | torch.LongTensor()
64 |
65 | pad = 1 if self.use_padding else 0
66 |
67 | self.enc1 = conv_block(inplanes, 64, 3, pad=pad, activation=self.activation, use_bn=self.use_batch_norm)
68 | self.enc2 = conv_block(64, 128, 3, pad=pad, activation=self.activation, use_bn=self.use_batch_norm)
69 | self.enc3 = conv_block(128, 256, 3, pad=pad, activation=self.activation, use_bn=self.use_batch_norm)
70 | self.enc4 = conv_block(256, 512, 3, pad=pad, activation=self.activation, use_bn=self.use_batch_norm)
71 | self.enc5 = conv_block(512, 1024, 3, pad=pad, activation=self.activation, use_bn=self.use_batch_norm)
72 |
73 | self.pool1 = pool_layer()
74 | self.pool2 = pool_layer()
75 | self.pool3 = pool_layer()
76 | self.pool4 = pool_layer()
77 |
78 | self.dec4 = conv_block(1024, 512, 3, pad=pad, activation=self.activation, use_bn=self.use_batch_norm)
79 | self.dec3 = conv_block(512, 256, 3, pad=pad, activation=self.activation, use_bn=self.use_batch_norm)
80 | self.dec2 = conv_block(256, 128, 3, pad=pad, activation=self.activation, use_bn=self.use_batch_norm)
81 | self.dec1 = conv_block(128, 64, 3, pad=pad, activation=self.activation, use_bn=self.use_batch_norm)
82 |
83 | self.upsample4 = upsample_layer(1024, 512)
84 | self.upsample3 = upsample_layer(512, 256)
85 | self.upsample2 = upsample_layer(256, 128)
86 | self.upsample1 = upsample_layer(128, 64)
87 |
88 | if self.use_padding:
89 | self.conv_final = nn.Sequential(
90 | nn.Conv2d(64, self.outplanes, kernel_size=1, stride=1),
91 | nn.ReLU())
92 | else:
93 | if not isinstance(target_size, tuple):
94 | target_size = tuple(target_size)
95 | self.conv_final = nn.Sequential(
96 | nn.Conv2d(64, self.outplanes, kernel_size=1, stride=1),
97 | nn.ReLU(),
98 | nn.UpsamplingBilinear2d(size=target_size))
99 |
100 | # Weight initialization
101 | for m in self.modules():
102 | if isinstance(m, nn.Conv2d) or isinstance(m, nn.ConvTranspose2d):
103 | n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
104 | m.weight.data.normal_(0, math.sqrt(2. / n))
105 | elif isinstance(m, nn.BatchNorm2d):
106 | m.weight.data.fill_(1)
107 | m.bias.data.zero_()
108 |
109 | def _print(self, x, tag=[]):
110 | if isinstance(tag, str):
111 | tag = [tag]
112 | if self.debug:
113 | if tag:
114 | print('%s: %s' % (' '.join(filter(None, tag)), x.size()))
115 | else:
116 | print(x.size())
117 |
118 | def _crop_and_concat(self, upsampled, bypass, crop=False, tag=''):
119 | self._print(bypass, [tag, 'bypass'])
120 | self._print(upsampled, [tag, 'upsampled'])
121 | if crop:
122 | c = (bypass.size()[2] - upsampled.size()[2]) // 2
123 | bypass = F.pad(bypass, (-c, -c, -c, -c))
124 | return torch.cat((upsampled, bypass), 1)
125 |
126 | def forward(self, x):
127 |
128 | enc1_out = self.enc1(x) # 64
129 | self._print(enc1_out, 'enc1')
130 | enc2_out = self.enc2(self.pool1(enc1_out)) # 128
131 | self._print(enc2_out, 'enc2')
132 | enc3_out = self.enc3(self.pool2(enc2_out)) # 256
133 | self._print(enc3_out, 'enc3')
134 | enc4_out = self.enc4(self.pool3(enc3_out)) # 512
135 | self._print(enc4_out, 'enc4')
136 | enc5_out = self.enc5(self.pool4(enc4_out)) # 1024
137 | self._print(enc5_out, 'enc5')
138 |
139 | crop = False if self.use_padding else True
140 | dec4_out = self.dec4(self._crop_and_concat(self.upsample4(enc5_out), enc4_out, crop, 'dec4'))
141 | dec3_out = self.dec3(self._crop_and_concat(self.upsample3(dec4_out), enc3_out, crop, 'dec3'))
142 | dec2_out = self.dec2(self._crop_and_concat(self.upsample2(dec3_out), enc2_out, crop, 'dec2'))
143 | dec1_out = self.dec1(self._crop_and_concat(self.upsample1(dec2_out), enc1_out, crop, 'dec1'))
144 | conv_final_out = self.conv_final(dec1_out)
145 | self._print(conv_final_out, 'final')
146 |
147 | return conv_final_out
148 |
149 | def name(self):
150 | return 'cnet'
151 |
--------------------------------------------------------------------------------
/models/model_countception.py:
--------------------------------------------------------------------------------
1 | """ Counception Model
2 | A Pytorch implementation of Count-ception
3 |
4 | Inspired by: https://arxiv.org/abs/1703.08710
5 | """
6 | import torch
7 | import torch.nn as nn
8 | import torch.nn.init as init
9 |
10 |
11 | class ConvBlock(nn.Module):
12 | def __init__(self, in_chan, out_chan, ksize=3, stride=1, pad=0, activation=nn.LeakyReLU()):
13 | super(ConvBlock, self).__init__()
14 | self.conv1 = nn.Conv2d(in_chan, out_chan, kernel_size=ksize, stride=stride, padding=pad)
15 | self.activation = activation
16 | self.batch_norm = nn.BatchNorm2d(out_chan)
17 |
18 | def forward(self, x):
19 | return self.activation(self.batch_norm(self.conv1(x)))
20 |
21 |
22 | class SimpleBlock(nn.Module):
23 | def __init__(self, in_chan, out_chan_1x1, out_chan_3x3, activation=nn.LeakyReLU()):
24 | super(SimpleBlock, self).__init__()
25 | self.conv1 = ConvBlock(in_chan, out_chan_1x1, ksize=1, pad=0, activation=activation)
26 | self.conv2 = ConvBlock(in_chan, out_chan_3x3, ksize=3, pad=1, activation=activation)
27 |
28 | def forward(self, x):
29 | conv1_out = self.conv1(x)
30 | conv2_out = self.conv2(x)
31 | output = torch.cat([conv1_out, conv2_out], 1)
32 | return output
33 |
34 |
35 | class ModelCountception(nn.Module):
36 | def __init__(self, inplanes=3, outplanes=1, use_logits=False, logits_per_output=12, debug=False):
37 | super(ModelCountception, self).__init__()
38 | # params
39 | self.inplanes = inplanes
40 | self.outplanes = outplanes
41 | self.activation = nn.LeakyReLU(0.01)
42 | self.final_activation = nn.LeakyReLU(0.01)
43 | self.patch_size = 32
44 | self.use_logits = use_logits
45 | self.logits_per_output = logits_per_output
46 | self.debug = debug
47 |
48 | torch.LongTensor()
49 |
50 | self.conv1 = ConvBlock(self.inplanes, 64, ksize=3, pad=self.patch_size, activation=self.activation)
51 | self.simple1 = SimpleBlock(64, 16, 16, activation=self.activation)
52 | self.simple2 = SimpleBlock(32, 16, 32, activation=self.activation)
53 | self.conv2 = ConvBlock(48, 16, ksize=14, activation=self.activation)
54 | self.simple3 = SimpleBlock(16, 112, 48, activation=self.activation)
55 | self.simple4 = SimpleBlock(160, 64, 32, activation=self.activation)
56 | self.simple5 = SimpleBlock(96, 40, 40, activation=self.activation)
57 | self.simple6 = SimpleBlock(80, 32, 96, activation=self.activation)
58 | self.conv3 = ConvBlock(128, 32, ksize=20, activation=self.activation)
59 | self.conv4 = ConvBlock(32, 64, ksize=1, activation=self.activation)
60 | self.conv5 = ConvBlock(64, 64, ksize=1, activation=self.activation)
61 | if use_logits:
62 | self.conv6 = nn.ModuleList([ConvBlock(
63 | 64, logits_per_output, ksize=1, activation=self.final_activation) for _ in range(outplanes)])
64 | else:
65 | self.conv6 = ConvBlock(64, self.outplanes, ksize=1, activation=self.final_activation)
66 |
67 | # Weight initialization
68 | for m in self.modules():
69 | if isinstance(m, nn.Conv2d) or isinstance(m, nn.ConvTranspose2d):
70 | init.xavier_uniform(m.weight, gain=init.calculate_gain('leaky_relu', param=0.01))
71 | elif isinstance(m, nn.BatchNorm2d):
72 | m.weight.data.fill_(1)
73 | m.bias.data.zero_()
74 |
75 | def _print(self, x):
76 | if self.debug:
77 | print(x.size())
78 |
79 | def forward(self, x):
80 | net = self.conv1(x) # 32
81 | self._print(net)
82 | net = self.simple1(net)
83 | self._print(net)
84 | net = self.simple2(net)
85 | self._print(net)
86 | net = self.conv2(net)
87 | self._print(net)
88 | net = self.simple3(net)
89 | self._print(net)
90 | net = self.simple4(net)
91 | self._print(net)
92 | net = self.simple5(net)
93 | self._print(net)
94 | net = self.simple6(net)
95 | self._print(net)
96 | net = self.conv3(net)
97 | self._print(net)
98 | net = self.conv4(net)
99 | self._print(net)
100 | net = self.conv5(net)
101 | self._print(net)
102 | if self.use_logits:
103 | net = [c(net) for c in self.conv6]
104 | [self._print(n) for n in net]
105 | else:
106 | net = self.conv6(net)
107 | self._print(net)
108 | return net
109 |
110 | def name(self):
111 | return 'countception'
112 |
--------------------------------------------------------------------------------
/mytransforms.py:
--------------------------------------------------------------------------------
1 | """ A few fb.resnet.torch like tranforms
2 | Most taken from https://github.com/pytorch/vision/pull/27
3 | """
4 | import torch
5 | import random
6 | import cv2
7 | import numpy as np
8 |
9 |
10 |
11 |
12 | class Grayscale(object):
13 |
14 | def __call__(self, img):
15 | gs = img.clone()
16 | gs[0].mul_(0.299).add_(0.587, gs[1]).add_(0.114, gs[2])
17 | gs[1].copy_(gs[0])
18 | gs[2].copy_(gs[0])
19 | return gs
20 |
21 |
22 | class Saturation(object):
23 |
24 | def __init__(self, var):
25 | self.var = var
26 |
27 | def __call__(self, img):
28 | gs = Grayscale()(img)
29 | alpha = random.uniform(0, self.var)
30 | return img.lerp(gs, alpha)
31 |
32 |
33 | class Brightness(object):
34 |
35 | def __init__(self, var):
36 | self.var = var
37 |
38 | def __call__(self, img):
39 | gs = img.new().resize_as_(img).zero_()
40 | alpha = random.uniform(0, self.var)
41 | return img.lerp(gs, alpha)
42 |
43 |
44 | class Contrast(object):
45 |
46 | def __init__(self, var):
47 | self.var = var
48 |
49 | def __call__(self, img):
50 | gs = Grayscale()(img)
51 | gs.fill_(gs.mean())
52 | alpha = random.uniform(0, self.var)
53 | return img.lerp(gs, alpha)
54 |
55 |
56 | class RandomOrder(object):
57 | """ Composes several transforms together in random order.
58 | """
59 |
60 | def __init__(self, transforms):
61 | self.transforms = transforms
62 |
63 | def __call__(self, img):
64 | if self.transforms is None:
65 | return img
66 | order = torch.randperm(len(self.transforms))
67 | for i in order:
68 | img = self.transforms[i](img)
69 | return img
70 |
71 |
72 | class ColorJitter(RandomOrder):
73 |
74 | def __init__(self, brightness=0.4, contrast=0.4, saturation=0.4):
75 | self.transforms = []
76 | if brightness != 0:
77 | self.transforms.append(Brightness(brightness))
78 | if contrast != 0:
79 | self.transforms.append(Contrast(contrast))
80 | if saturation != 0:
81 | self.transforms.append(Saturation(saturation))
82 |
83 |
84 | class NormalizeImg:
85 | """Normalize each image or patch by its own mean/std
86 | """
87 |
88 | def __init__(self, std_epsilon=.0001):
89 | self.std_epsilon = std_epsilon
90 |
91 | def __call__(self, img):
92 | # This should still be a H x W x C Numpy/OpenCv compat image, not a Torch Tensor
93 | assert isinstance(img, np.ndarray)
94 | mean, std = cv2.meanStdDev(img)
95 | mean, std = mean.astype(np.float32), std.astype(np.float32)
96 | img = img.astype(np.float32)
97 | img = (img - np.squeeze(mean)) / (np.squeeze(std) + self.std_epsilon)
98 | return img
99 |
100 |
101 | class ToTensor:
102 | def __call__(self, img):
103 | assert isinstance(img, np.ndarray)
104 | # handle numpy array
105 | img = torch.from_numpy(img.transpose((2, 0, 1)))
106 | if isinstance(img, torch.ByteTensor):
107 | return img.float().div(255)
108 | else:
109 | return img
110 |
--------------------------------------------------------------------------------
/scripts/other/README.md:
--------------------------------------------------------------------------------
1 | The scripts in this folder are work by:
2 | * notebook -- https://www.kaggle.com/radustoicescu
3 | * script.py -- https://www.kaggle.com/threeplusone
--------------------------------------------------------------------------------
/scripts/other/__notebook__.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {
6 | "_cell_guid": "7daa5a37-ea6b-2274-4bc1-fadb0a2fc41a"
7 | },
8 | "source": [
9 | "### Get dot coordinates using blob_log from skimage library"
10 | ]
11 | },
12 | {
13 | "cell_type": "code",
14 | "execution_count": 1,
15 | "metadata": {
16 | "_cell_guid": "35cabbd8-8e7b-5be7-bd00-61b1addc2d75",
17 | "collapsed": true
18 | },
19 | "outputs": [],
20 | "source": [
21 | "import numpy as np\n",
22 | "import pandas as pd\n",
23 | "import os\n",
24 | "import cv2\n",
25 | "import matplotlib.pyplot as plt\n",
26 | "import skimage.feature\n",
27 | "%matplotlib inline"
28 | ]
29 | },
30 | {
31 | "cell_type": "code",
32 | "execution_count": 2,
33 | "metadata": {
34 | "_cell_guid": "3a6329c7-aa3b-b576-b6e1-d675e6e8fbf2",
35 | "collapsed": false
36 | },
37 | "outputs": [],
38 | "source": [
39 | "classes = [\"adult_males\", \"subadult_males\", \"adult_females\", \"juveniles\", \"pups\", \"error\"]\n",
40 | "coords_cols = [\"filename\", \"x\", \"y\", \"category\"]\n",
41 | "\n",
42 | "file_names = os.listdir(\"/data/x/sealion/Train\")\n",
43 | "file_names = sorted(file_names, key=lambda \n",
44 | " item: (int(item.partition('.')[0]) if item[0].isdigit() else float('inf'), item)) \n",
45 | "\n",
46 | "indices = [531, 946, 34, 30, 290, 406, 380, 913, 621, 811, 7, 421, 292, 66, 593, 490, 909, 800, 215, \n",
47 | " 426, 475, 614, 184, 905, 97, 882, 776, 899, 344, 473, 510, 234, 291, 331, 433, 712, 741, 767, 912]\n",
48 | "#indices = [290, 291]\n",
49 | "# select a subset of files to run on\n",
50 | "file_names = [file_names[i] for i in indices]"
51 | ]
52 | },
53 | {
54 | "cell_type": "code",
55 | "execution_count": 17,
56 | "metadata": {
57 | "_cell_guid": "7755c681-04df-368a-aca2-f099dd9ce805",
58 | "collapsed": false
59 | },
60 | "outputs": [
61 | {
62 | "name": "stdout",
63 | "output_type": "stream",
64 | "text": [
65 | "531.jpg\n"
66 | ]
67 | },
68 | {
69 | "name": "stdout",
70 | "output_type": "stream",
71 | "text": [
72 | "163.097349682\nWarning: Bad data for 531.jpg\n946.jpg\n"
73 | ]
74 | },
75 | {
76 | "name": "stdout",
77 | "output_type": "stream",
78 | "text": [
79 | "113.417159824\nWarning: Bad data for 946.jpg\n34.jpg\n"
80 | ]
81 | },
82 | {
83 | "name": "stdout",
84 | "output_type": "stream",
85 | "text": [
86 | "154.315978785\nWarning: Bad data for 34.jpg\n30.jpg\n"
87 | ]
88 | },
89 | {
90 | "name": "stdout",
91 | "output_type": "stream",
92 | "text": [
93 | "31.2134717832\n"
94 | ]
95 | },
96 | {
97 | "name": "stdout",
98 | "output_type": "stream",
99 | "text": [
100 | "290.jpg\n"
101 | ]
102 | },
103 | {
104 | "name": "stdout",
105 | "output_type": "stream",
106 | "text": [
107 | "173.808650511\nWarning: Bad data for 290.jpg\n406.jpg\n"
108 | ]
109 | },
110 | {
111 | "name": "stdout",
112 | "output_type": "stream",
113 | "text": [
114 | "67.2546158589\nWarning: Bad data for 406.jpg\n380.jpg\n"
115 | ]
116 | },
117 | {
118 | "name": "stdout",
119 | "output_type": "stream",
120 | "text": [
121 | "33.9865956805\n"
122 | ]
123 | },
124 | {
125 | "name": "stdout",
126 | "output_type": "stream",
127 | "text": [
128 | "913.jpg\n"
129 | ]
130 | },
131 | {
132 | "name": "stdout",
133 | "output_type": "stream",
134 | "text": [
135 | "25.8726279355\n"
136 | ]
137 | },
138 | {
139 | "name": "stdout",
140 | "output_type": "stream",
141 | "text": [
142 | "621.jpg\n"
143 | ]
144 | },
145 | {
146 | "name": "stdout",
147 | "output_type": "stream",
148 | "text": [
149 | "141.90937851\nWarning: Bad data for 621.jpg\n811.jpg\n"
150 | ]
151 | },
152 | {
153 | "name": "stdout",
154 | "output_type": "stream",
155 | "text": [
156 | "98.1498576208\nWarning: Bad data for 811.jpg\n7.jpg\n"
157 | ]
158 | },
159 | {
160 | "name": "stdout",
161 | "output_type": "stream",
162 | "text": [
163 | "102.454059496\nWarning: Bad data for 7.jpg\n421.jpg\n"
164 | ]
165 | },
166 | {
167 | "name": "stdout",
168 | "output_type": "stream",
169 | "text": [
170 | "115.55434783\nWarning: Bad data for 421.jpg\n292.jpg\n"
171 | ]
172 | },
173 | {
174 | "name": "stdout",
175 | "output_type": "stream",
176 | "text": [
177 | "11.0090252666\n"
178 | ]
179 | },
180 | {
181 | "name": "stdout",
182 | "output_type": "stream",
183 | "text": [
184 | "66.jpg\n"
185 | ]
186 | },
187 | {
188 | "name": "stdout",
189 | "output_type": "stream",
190 | "text": [
191 | "10.0156795507\n"
192 | ]
193 | },
194 | {
195 | "name": "stdout",
196 | "output_type": "stream",
197 | "text": [
198 | "593.jpg\n"
199 | ]
200 | },
201 | {
202 | "name": "stdout",
203 | "output_type": "stream",
204 | "text": [
205 | "6.46708742401\n"
206 | ]
207 | },
208 | {
209 | "name": "stdout",
210 | "output_type": "stream",
211 | "text": [
212 | "490.jpg\n"
213 | ]
214 | },
215 | {
216 | "name": "stdout",
217 | "output_type": "stream",
218 | "text": [
219 | "156.886961145\nWarning: Bad data for 490.jpg\n909.jpg\n"
220 | ]
221 | },
222 | {
223 | "name": "stdout",
224 | "output_type": "stream",
225 | "text": [
226 | "93.3790219399\nWarning: Bad data for 909.jpg\n800.jpg\n"
227 | ]
228 | },
229 | {
230 | "name": "stdout",
231 | "output_type": "stream",
232 | "text": [
233 | "156.377165604\nWarning: Bad data for 800.jpg\n215.jpg\n"
234 | ]
235 | },
236 | {
237 | "name": "stdout",
238 | "output_type": "stream",
239 | "text": [
240 | "120.281212143\nWarning: Bad data for 215.jpg\n426.jpg\n"
241 | ]
242 | },
243 | {
244 | "name": "stdout",
245 | "output_type": "stream",
246 | "text": [
247 | "4.59449292657\n"
248 | ]
249 | },
250 | {
251 | "name": "stdout",
252 | "output_type": "stream",
253 | "text": [
254 | "475.jpg\n"
255 | ]
256 | },
257 | {
258 | "name": "stdout",
259 | "output_type": "stream",
260 | "text": [
261 | "27.8335969079\n"
262 | ]
263 | },
264 | {
265 | "name": "stdout",
266 | "output_type": "stream",
267 | "text": [
268 | "614.jpg\n"
269 | ]
270 | },
271 | {
272 | "name": "stdout",
273 | "output_type": "stream",
274 | "text": [
275 | "86.4739226162\nWarning: Bad data for 614.jpg\n184.jpg\n"
276 | ]
277 | },
278 | {
279 | "name": "stdout",
280 | "output_type": "stream",
281 | "text": [
282 | "92.231638618\nWarning: Bad data for 184.jpg\n905.jpg\n"
283 | ]
284 | },
285 | {
286 | "name": "stdout",
287 | "output_type": "stream",
288 | "text": [
289 | "146.463161792\nWarning: Bad data for 905.jpg\n97.jpg\n"
290 | ]
291 | },
292 | {
293 | "name": "stdout",
294 | "output_type": "stream",
295 | "text": [
296 | "7.23316007535\n"
297 | ]
298 | },
299 | {
300 | "name": "stdout",
301 | "output_type": "stream",
302 | "text": [
303 | "882.jpg\n"
304 | ]
305 | },
306 | {
307 | "name": "stdout",
308 | "output_type": "stream",
309 | "text": [
310 | "62.3129591392\nWarning: Bad data for 882.jpg\n776.jpg\n"
311 | ]
312 | },
313 | {
314 | "name": "stdout",
315 | "output_type": "stream",
316 | "text": [
317 | "4.32146405759\n"
318 | ]
319 | },
320 | {
321 | "name": "stdout",
322 | "output_type": "stream",
323 | "text": [
324 | "899.jpg\n"
325 | ]
326 | },
327 | {
328 | "name": "stdout",
329 | "output_type": "stream",
330 | "text": [
331 | "6.21177445166\n"
332 | ]
333 | },
334 | {
335 | "name": "stdout",
336 | "output_type": "stream",
337 | "text": [
338 | "344.jpg\n"
339 | ]
340 | },
341 | {
342 | "name": "stdout",
343 | "output_type": "stream",
344 | "text": [
345 | "65.4390775954\nWarning: Bad data for 344.jpg\n473.jpg\n"
346 | ]
347 | },
348 | {
349 | "name": "stdout",
350 | "output_type": "stream",
351 | "text": [
352 | "13.8186259674\n"
353 | ]
354 | },
355 | {
356 | "name": "stdout",
357 | "output_type": "stream",
358 | "text": [
359 | "510.jpg\n"
360 | ]
361 | },
362 | {
363 | "name": "stdout",
364 | "output_type": "stream",
365 | "text": [
366 | "8.1044958258\n"
367 | ]
368 | },
369 | {
370 | "name": "stdout",
371 | "output_type": "stream",
372 | "text": [
373 | "234.jpg\n"
374 | ]
375 | },
376 | {
377 | "name": "stdout",
378 | "output_type": "stream",
379 | "text": [
380 | "123.899784005\nWarning: Bad data for 234.jpg\n291.jpg\n"
381 | ]
382 | },
383 | {
384 | "name": "stdout",
385 | "output_type": "stream",
386 | "text": [
387 | "8.37217530004\n"
388 | ]
389 | },
390 | {
391 | "name": "stdout",
392 | "output_type": "stream",
393 | "text": [
394 | "331.jpg\n"
395 | ]
396 | },
397 | {
398 | "name": "stdout",
399 | "output_type": "stream",
400 | "text": [
401 | "4.38011716182\n"
402 | ]
403 | },
404 | {
405 | "name": "stdout",
406 | "output_type": "stream",
407 | "text": [
408 | "433.jpg\n"
409 | ]
410 | },
411 | {
412 | "name": "stdout",
413 | "output_type": "stream",
414 | "text": [
415 | "8.13684525821\n"
416 | ]
417 | },
418 | {
419 | "name": "stdout",
420 | "output_type": "stream",
421 | "text": [
422 | "712.jpg\n"
423 | ]
424 | },
425 | {
426 | "name": "stdout",
427 | "output_type": "stream",
428 | "text": [
429 | "121.709471527\nWarning: Bad data for 712.jpg\n741.jpg\n"
430 | ]
431 | },
432 | {
433 | "name": "stdout",
434 | "output_type": "stream",
435 | "text": [
436 | "6.15084790713\n"
437 | ]
438 | },
439 | {
440 | "name": "stdout",
441 | "output_type": "stream",
442 | "text": [
443 | "767.jpg\n"
444 | ]
445 | },
446 | {
447 | "name": "stdout",
448 | "output_type": "stream",
449 | "text": [
450 | "163.374341254\nWarning: Bad data for 767.jpg\n912.jpg\n"
451 | ]
452 | },
453 | {
454 | "name": "stdout",
455 | "output_type": "stream",
456 | "text": [
457 | "5.23696682721\n"
458 | ]
459 | }
460 | ],
461 | "source": [
462 | "# dataframe to store results in\n",
463 | "count_df = pd.DataFrame(index=file_names, columns=classes).fillna(0)\n",
464 | "coords = []\n",
465 | "\n",
466 | "for filename in file_names:\n",
467 | " print(filename)\n",
468 | " \n",
469 | " # read the Train and Train Dotted images\n",
470 | " image_1 = cv2.imread(\"/data/x/sealion/TrainDotted/\" + filename)\n",
471 | " image_2 = cv2.imread(\"/data/x/sealion/Train/\" + filename)\n",
472 | " \n",
473 | " # absolute difference between Train and Train Dotted\n",
474 | " image_3 = cv2.absdiff(image_1, image_2)\n",
475 | " \n",
476 | " # mask out blackened regions from Train Dotted\n",
477 | " mask_1 = cv2.cvtColor(image_1, cv2.COLOR_BGR2GRAY)\n",
478 | " mask_1[mask_1 < 10] = 0\n",
479 | " mask_1[mask_1 > 0] = 255\n",
480 | " \n",
481 | " mask_2 = cv2.cvtColor(image_2, cv2.COLOR_BGR2GRAY)\n",
482 | " mask_2[mask_2 < 10] = 0\n",
483 | " mask_2[mask_2 > 0] = 255\n",
484 | " \n",
485 | " image_4 = cv2.bitwise_or(image_3, image_3, mask=mask_1)\n",
486 | " \n",
487 | " # Detect bad data. If train and dotted images are very different then somethings wrong.\n",
488 | " avg_diff = image_4.sum() / (image_4.shape[0] * image_4.shape[1])\n",
489 | " print(avg_diff)\n",
490 | " if avg_diff > 60:\n",
491 | " print('Warning: Bad data for %s' % filename)\n",
492 | " continue \n",
493 | " \n",
494 | " image_5 = cv2.bitwise_or(image_4, image_4, mask=mask_2) \n",
495 | " \n",
496 | " # convert to grayscale to be accepted by skimage.feature.blob_log\n",
497 | " image_6 = cv2.cvtColor(image_5, cv2.COLOR_BGR2GRAY)\n",
498 | " \n",
499 | " # detect blobs\n",
500 | " blobs = skimage.feature.blob_log(image_6, min_sigma=3, max_sigma=4, num_sigma=2, threshold=0.02)\n",
501 | " \n",
502 | " # prepare the image to plot the results on\n",
503 | " image_7 = cv2.cvtColor(image_6, cv2.COLOR_GRAY2BGR)\n",
504 | " \n",
505 | " sizes = [list()] * 6\n",
506 | " for blob in blobs:\n",
507 | " # get the coordinates for each blob\n",
508 | " y, x, s = blob\n",
509 | " # get the color of the pixel from Train Dotted in the center of the blob\n",
510 | " b, g, r = image_1[int(y)][int(x)][:]\n",
511 | " \n",
512 | " # decision tree to pick the class of the blob by looking at the color in Train Dotted\n",
513 | " class_idx = -1\n",
514 | " if r > 200 and b < 50 and g < 50: # RED\n",
515 | " class_idx = 0\n",
516 | " cv2.circle(image_7, (int(x), int(y)), 8, (0, 0, 255), 2) \n",
517 | " elif r > 200 and b > 200 and g < 50: # MAGENTA\n",
518 | " class_idx = 1\n",
519 | " cv2.circle(image_7, (int(x), int(y)), 8, (250, 10, 250), 2) \n",
520 | " elif r < 100 and b < 100 and 150 < g < 200: # GREEN\n",
521 | " class_idx = 4\n",
522 | " cv2.circle(image_7, (int(x), int(y)), 8, (20, 180, 35), 2) \n",
523 | " elif r < 100 and 100 < b and g < 100: # BLUE\n",
524 | " class_idx = 3\n",
525 | " cv2.circle(image_7, (int(x), int(y)), 8, (180, 60, 30), 2)\n",
526 | " elif r < 150 and b < 50 and g < 100: # BROWN\n",
527 | " class_idx = 2\n",
528 | " cv2.circle(image_7, (int(x), int(y)), 8, (0, 42, 84), 2) \n",
529 | " else:\n",
530 | " class_idx = 5\n",
531 | " cv2.circle(image_7, (int(x), int(y)), 8, (255, 255, 155), 2)\n",
532 | " \n",
533 | " sizes[class_idx].append(s)\n",
534 | " class_name = classes[class_idx]\n",
535 | " count_df[class_name][filename] += 1\n",
536 | " record = dict(filename=filename, x=x, y=y, category=cls, )\n",
537 | " if -1 < class_idx < 5:\n",
538 | " coords.append(record)\n",
539 | " coords_df = pd.DataFrame.from_records(coords, columns=coords_cols)\n",
540 | " coords_df.x = coords_df.x.astype('int')\n",
541 | " coords_df.y = coords_df.y.astype('int')\n",
542 | " # output the results\n",
543 | " \n",
544 | "# f, ax = plt.subplots(3, 2, figsize=(10,16))\n",
545 | "# (ax1, ax2, ax3, ax4, ax5, ax6) = ax.flatten()\n",
546 | "# plt.title('%s'%filename)\n",
547 | " \n",
548 | "# ax1.imshow(cv2.cvtColor(image_2[:,:,:], cv2.COLOR_BGR2RGB))\n",
549 | "# ax1.set_title('Train')\n",
550 | "# ax2.imshow(cv2.cvtColor(image_1[:,:,:], cv2.COLOR_BGR2RGB))\n",
551 | "# ax2.set_title('Train Dotted')\n",
552 | "# ax3.imshow(cv2.cvtColor(image_3[:,:,:], cv2.COLOR_BGR2RGB))\n",
553 | "# ax3.set_title('Train Dotted - Train')\n",
554 | "# ax4.imshow(cv2.cvtColor(image_5[:,:,:], cv2.COLOR_BGR2RGB))\n",
555 | "# ax4.set_title('Mask blackened areas of Train Dotted')\n",
556 | "# ax5.imshow(image_6[:,:], cmap='gray')\n",
557 | "# ax5.set_title('Grayscale for input to blob_log')\n",
558 | "# ax6.imshow(cv2.cvtColor(image_7[:,:,:], cv2.COLOR_BGR2RGB))\n",
559 | "# ax6.set_title('Result')\n",
560 | "\n",
561 | "# plt.show()"
562 | ]
563 | },
564 | {
565 | "cell_type": "markdown",
566 | "metadata": {
567 | "_cell_guid": "d03424c1-b12b-ae53-2fed-1dff86398164"
568 | },
569 | "source": [
570 | "### Check count results"
571 | ]
572 | },
573 | {
574 | "cell_type": "code",
575 | "execution_count": 18,
576 | "metadata": {
577 | "_cell_guid": "c9549615-3a64-2ef2-2be0-1946e07c2ee2",
578 | "collapsed": false
579 | },
580 | "outputs": [
581 | {
582 | "data": {
583 | "text/html": [
584 | "\n",
585 | "
\n",
586 | " \n",
587 | " \n",
588 | " | \n",
589 | " adult_males | \n",
590 | " subadult_males | \n",
591 | " adult_females | \n",
592 | " juveniles | \n",
593 | " pups | \n",
594 | " error | \n",
595 | "
\n",
596 | " \n",
597 | " \n",
598 | " \n",
599 | " | 531.jpg | \n",
600 | " 0 | \n",
601 | " 0 | \n",
602 | " 0 | \n",
603 | " 0 | \n",
604 | " 0 | \n",
605 | " 0 | \n",
606 | "
\n",
607 | " \n",
608 | " | 946.jpg | \n",
609 | " 0 | \n",
610 | " 0 | \n",
611 | " 0 | \n",
612 | " 0 | \n",
613 | " 0 | \n",
614 | " 0 | \n",
615 | "
\n",
616 | " \n",
617 | " | 34.jpg | \n",
618 | " 0 | \n",
619 | " 0 | \n",
620 | " 0 | \n",
621 | " 0 | \n",
622 | " 0 | \n",
623 | " 0 | \n",
624 | "
\n",
625 | " \n",
626 | " | 30.jpg | \n",
627 | " 2 | \n",
628 | " 0 | \n",
629 | " 3 | \n",
630 | " 3 | \n",
631 | " 0 | \n",
632 | " 1 | \n",
633 | "
\n",
634 | " \n",
635 | " | 290.jpg | \n",
636 | " 0 | \n",
637 | " 0 | \n",
638 | " 0 | \n",
639 | " 0 | \n",
640 | " 0 | \n",
641 | " 0 | \n",
642 | "
\n",
643 | " \n",
644 | " | 406.jpg | \n",
645 | " 0 | \n",
646 | " 0 | \n",
647 | " 0 | \n",
648 | " 0 | \n",
649 | " 0 | \n",
650 | " 0 | \n",
651 | "
\n",
652 | " \n",
653 | " | 380.jpg | \n",
654 | " 2 | \n",
655 | " 0 | \n",
656 | " 111 | \n",
657 | " 46 | \n",
658 | " 0 | \n",
659 | " 150 | \n",
660 | "
\n",
661 | " \n",
662 | " | 913.jpg | \n",
663 | " 1 | \n",
664 | " 0 | \n",
665 | " 109 | \n",
666 | " 33 | \n",
667 | " 0 | \n",
668 | " 47 | \n",
669 | "
\n",
670 | " \n",
671 | " | 621.jpg | \n",
672 | " 0 | \n",
673 | " 0 | \n",
674 | " 0 | \n",
675 | " 0 | \n",
676 | " 0 | \n",
677 | " 0 | \n",
678 | "
\n",
679 | " \n",
680 | " | 811.jpg | \n",
681 | " 0 | \n",
682 | " 0 | \n",
683 | " 0 | \n",
684 | " 0 | \n",
685 | " 0 | \n",
686 | " 0 | \n",
687 | "
\n",
688 | " \n",
689 | " | 7.jpg | \n",
690 | " 0 | \n",
691 | " 0 | \n",
692 | " 0 | \n",
693 | " 0 | \n",
694 | " 0 | \n",
695 | " 0 | \n",
696 | "
\n",
697 | " \n",
698 | " | 421.jpg | \n",
699 | " 0 | \n",
700 | " 0 | \n",
701 | " 0 | \n",
702 | " 0 | \n",
703 | " 0 | \n",
704 | " 0 | \n",
705 | "
\n",
706 | " \n",
707 | " | 292.jpg | \n",
708 | " 5 | \n",
709 | " 5 | \n",
710 | " 70 | \n",
711 | " 43 | \n",
712 | " 1 | \n",
713 | " 17 | \n",
714 | "
\n",
715 | " \n",
716 | " | 66.jpg | \n",
717 | " 8 | \n",
718 | " 5 | \n",
719 | " 61 | \n",
720 | " 17 | \n",
721 | " 2 | \n",
722 | " 32 | \n",
723 | "
\n",
724 | " \n",
725 | " | 593.jpg | \n",
726 | " 1 | \n",
727 | " 2 | \n",
728 | " 32 | \n",
729 | " 58 | \n",
730 | " 0 | \n",
731 | " 3 | \n",
732 | "
\n",
733 | " \n",
734 | " | 490.jpg | \n",
735 | " 0 | \n",
736 | " 0 | \n",
737 | " 0 | \n",
738 | " 0 | \n",
739 | " 0 | \n",
740 | " 0 | \n",
741 | "
\n",
742 | " \n",
743 | " | 909.jpg | \n",
744 | " 0 | \n",
745 | " 0 | \n",
746 | " 0 | \n",
747 | " 0 | \n",
748 | " 0 | \n",
749 | " 0 | \n",
750 | "
\n",
751 | " \n",
752 | " | 800.jpg | \n",
753 | " 0 | \n",
754 | " 0 | \n",
755 | " 0 | \n",
756 | " 0 | \n",
757 | " 0 | \n",
758 | " 0 | \n",
759 | "
\n",
760 | " \n",
761 | " | 215.jpg | \n",
762 | " 0 | \n",
763 | " 0 | \n",
764 | " 0 | \n",
765 | " 0 | \n",
766 | " 0 | \n",
767 | " 0 | \n",
768 | "
\n",
769 | " \n",
770 | " | 426.jpg | \n",
771 | " 2 | \n",
772 | " 6 | \n",
773 | " 34 | \n",
774 | " 42 | \n",
775 | " 5 | \n",
776 | " 1 | \n",
777 | "
\n",
778 | " \n",
779 | " | 475.jpg | \n",
780 | " 3 | \n",
781 | " 6 | \n",
782 | " 61 | \n",
783 | " 85 | \n",
784 | " 0 | \n",
785 | " 5 | \n",
786 | "
\n",
787 | " \n",
788 | " | 614.jpg | \n",
789 | " 0 | \n",
790 | " 0 | \n",
791 | " 0 | \n",
792 | " 0 | \n",
793 | " 0 | \n",
794 | " 0 | \n",
795 | "
\n",
796 | " \n",
797 | " | 184.jpg | \n",
798 | " 0 | \n",
799 | " 0 | \n",
800 | " 0 | \n",
801 | " 0 | \n",
802 | " 0 | \n",
803 | " 0 | \n",
804 | "
\n",
805 | " \n",
806 | " | 905.jpg | \n",
807 | " 0 | \n",
808 | " 0 | \n",
809 | " 0 | \n",
810 | " 0 | \n",
811 | " 0 | \n",
812 | " 0 | \n",
813 | "
\n",
814 | " \n",
815 | " | 97.jpg | \n",
816 | " 0 | \n",
817 | " 0 | \n",
818 | " 27 | \n",
819 | " 11 | \n",
820 | " 0 | \n",
821 | " 0 | \n",
822 | "
\n",
823 | " \n",
824 | " | 882.jpg | \n",
825 | " 0 | \n",
826 | " 0 | \n",
827 | " 0 | \n",
828 | " 0 | \n",
829 | " 0 | \n",
830 | " 0 | \n",
831 | "
\n",
832 | " \n",
833 | " | 776.jpg | \n",
834 | " 8 | \n",
835 | " 2 | \n",
836 | " 51 | \n",
837 | " 2 | \n",
838 | " 29 | \n",
839 | " 6 | \n",
840 | "
\n",
841 | " \n",
842 | " | 899.jpg | \n",
843 | " 2 | \n",
844 | " 2 | \n",
845 | " 13 | \n",
846 | " 3 | \n",
847 | " 0 | \n",
848 | " 0 | \n",
849 | "
\n",
850 | " \n",
851 | " | 344.jpg | \n",
852 | " 0 | \n",
853 | " 0 | \n",
854 | " 0 | \n",
855 | " 0 | \n",
856 | " 0 | \n",
857 | " 0 | \n",
858 | "
\n",
859 | " \n",
860 | " | 473.jpg | \n",
861 | " 2 | \n",
862 | " 0 | \n",
863 | " 25 | \n",
864 | " 60 | \n",
865 | " 0 | \n",
866 | " 1 | \n",
867 | "
\n",
868 | " \n",
869 | " | 510.jpg | \n",
870 | " 5 | \n",
871 | " 1 | \n",
872 | " 6 | \n",
873 | " 0 | \n",
874 | " 0 | \n",
875 | " 0 | \n",
876 | "
\n",
877 | " \n",
878 | " | 234.jpg | \n",
879 | " 0 | \n",
880 | " 0 | \n",
881 | " 0 | \n",
882 | " 0 | \n",
883 | " 0 | \n",
884 | " 0 | \n",
885 | "
\n",
886 | " \n",
887 | " | 291.jpg | \n",
888 | " 43 | \n",
889 | " 20 | \n",
890 | " 299 | \n",
891 | " 24 | \n",
892 | " 237 | \n",
893 | " 11 | \n",
894 | "
\n",
895 | " \n",
896 | " | 331.jpg | \n",
897 | " 6 | \n",
898 | " 18 | \n",
899 | " 67 | \n",
900 | " 0 | \n",
901 | " 0 | \n",
902 | " 5 | \n",
903 | "
\n",
904 | " \n",
905 | " | 433.jpg | \n",
906 | " 0 | \n",
907 | " 3 | \n",
908 | " 9 | \n",
909 | " 43 | \n",
910 | " 0 | \n",
911 | " 3 | \n",
912 | "
\n",
913 | " \n",
914 | " | 712.jpg | \n",
915 | " 0 | \n",
916 | " 0 | \n",
917 | " 0 | \n",
918 | " 0 | \n",
919 | " 0 | \n",
920 | " 0 | \n",
921 | "
\n",
922 | " \n",
923 | " | 741.jpg | \n",
924 | " 34 | \n",
925 | " 5 | \n",
926 | " 231 | \n",
927 | " 38 | \n",
928 | " 176 | \n",
929 | " 1 | \n",
930 | "
\n",
931 | " \n",
932 | " | 767.jpg | \n",
933 | " 0 | \n",
934 | " 0 | \n",
935 | " 0 | \n",
936 | " 0 | \n",
937 | " 0 | \n",
938 | " 0 | \n",
939 | "
\n",
940 | " \n",
941 | " | 912.jpg | \n",
942 | " 30 | \n",
943 | " 2 | \n",
944 | " 251 | \n",
945 | " 45 | \n",
946 | " 205 | \n",
947 | " 432 | \n",
948 | "
\n",
949 | " \n",
950 | "
\n",
951 | "
\n",
955 | "
\n",
956 | " \n",
957 | " \n",
958 | " | \n",
959 | " adult_males | \n",
960 | " subadult_males | \n",
961 | " adult_females | \n",
962 | " juveniles | \n",
963 | " pups | \n",
964 | " error | \n",
965 | "
\n",
966 | " \n",
967 | " \n",
968 | " \n",
969 | " | 531.jpg | \n",
970 | " 0 | \n",
971 | " 0 | \n",
972 | " 0 | \n",
973 | " 0 | \n",
974 | " 0 | \n",
975 | " 0 | \n",
976 | "
\n",
977 | " \n",
978 | " | 946.jpg | \n",
979 | " 0 | \n",
980 | " 0 | \n",
981 | " 0 | \n",
982 | " 0 | \n",
983 | " 0 | \n",
984 | " 0 | \n",
985 | "
\n",
986 | " \n",
987 | " | 34.jpg | \n",
988 | " 0 | \n",
989 | " 0 | \n",
990 | " 0 | \n",
991 | " 0 | \n",
992 | " 0 | \n",
993 | " 0 | \n",
994 | "
\n",
995 | " \n",
996 | " | 30.jpg | \n",
997 | " 2 | \n",
998 | " 0 | \n",
999 | " 3 | \n",
1000 | " 3 | \n",
1001 | " 0 | \n",
1002 | " 1 | \n",
1003 | "
\n",
1004 | " \n",
1005 | " | 290.jpg | \n",
1006 | " 0 | \n",
1007 | " 0 | \n",
1008 | " 0 | \n",
1009 | " 0 | \n",
1010 | " 0 | \n",
1011 | " 0 | \n",
1012 | "
\n",
1013 | " \n",
1014 | " | 406.jpg | \n",
1015 | " 0 | \n",
1016 | " 0 | \n",
1017 | " 0 | \n",
1018 | " 0 | \n",
1019 | " 0 | \n",
1020 | " 0 | \n",
1021 | "
\n",
1022 | " \n",
1023 | " | 380.jpg | \n",
1024 | " 2 | \n",
1025 | " 0 | \n",
1026 | " 111 | \n",
1027 | " 46 | \n",
1028 | " 0 | \n",
1029 | " 150 | \n",
1030 | "
\n",
1031 | " \n",
1032 | " | 913.jpg | \n",
1033 | " 1 | \n",
1034 | " 0 | \n",
1035 | " 109 | \n",
1036 | " 33 | \n",
1037 | " 0 | \n",
1038 | " 47 | \n",
1039 | "
\n",
1040 | " \n",
1041 | " | 621.jpg | \n",
1042 | " 0 | \n",
1043 | " 0 | \n",
1044 | " 0 | \n",
1045 | " 0 | \n",
1046 | " 0 | \n",
1047 | " 0 | \n",
1048 | "
\n",
1049 | " \n",
1050 | " | 811.jpg | \n",
1051 | " 0 | \n",
1052 | " 0 | \n",
1053 | " 0 | \n",
1054 | " 0 | \n",
1055 | " 0 | \n",
1056 | " 0 | \n",
1057 | "
\n",
1058 | " \n",
1059 | " | 7.jpg | \n",
1060 | " 0 | \n",
1061 | " 0 | \n",
1062 | " 0 | \n",
1063 | " 0 | \n",
1064 | " 0 | \n",
1065 | " 0 | \n",
1066 | "
\n",
1067 | " \n",
1068 | " | 421.jpg | \n",
1069 | " 0 | \n",
1070 | " 0 | \n",
1071 | " 0 | \n",
1072 | " 0 | \n",
1073 | " 0 | \n",
1074 | " 0 | \n",
1075 | "
\n",
1076 | " \n",
1077 | " | 292.jpg | \n",
1078 | " 5 | \n",
1079 | " 5 | \n",
1080 | " 70 | \n",
1081 | " 43 | \n",
1082 | " 1 | \n",
1083 | " 17 | \n",
1084 | "
\n",
1085 | " \n",
1086 | " | 66.jpg | \n",
1087 | " 8 | \n",
1088 | " 5 | \n",
1089 | " 61 | \n",
1090 | " 17 | \n",
1091 | " 2 | \n",
1092 | " 32 | \n",
1093 | "
\n",
1094 | " \n",
1095 | " | 593.jpg | \n",
1096 | " 1 | \n",
1097 | " 2 | \n",
1098 | " 32 | \n",
1099 | " 58 | \n",
1100 | " 0 | \n",
1101 | " 3 | \n",
1102 | "
\n",
1103 | " \n",
1104 | " | 490.jpg | \n",
1105 | " 0 | \n",
1106 | " 0 | \n",
1107 | " 0 | \n",
1108 | " 0 | \n",
1109 | " 0 | \n",
1110 | " 0 | \n",
1111 | "
\n",
1112 | " \n",
1113 | " | 909.jpg | \n",
1114 | " 0 | \n",
1115 | " 0 | \n",
1116 | " 0 | \n",
1117 | " 0 | \n",
1118 | " 0 | \n",
1119 | " 0 | \n",
1120 | "
\n",
1121 | " \n",
1122 | " | 800.jpg | \n",
1123 | " 0 | \n",
1124 | " 0 | \n",
1125 | " 0 | \n",
1126 | " 0 | \n",
1127 | " 0 | \n",
1128 | " 0 | \n",
1129 | "
\n",
1130 | " \n",
1131 | " | 215.jpg | \n",
1132 | " 0 | \n",
1133 | " 0 | \n",
1134 | " 0 | \n",
1135 | " 0 | \n",
1136 | " 0 | \n",
1137 | " 0 | \n",
1138 | "
\n",
1139 | " \n",
1140 | " | 426.jpg | \n",
1141 | " 2 | \n",
1142 | " 6 | \n",
1143 | " 34 | \n",
1144 | " 42 | \n",
1145 | " 5 | \n",
1146 | " 1 | \n",
1147 | "
\n",
1148 | " \n",
1149 | " | 475.jpg | \n",
1150 | " 3 | \n",
1151 | " 6 | \n",
1152 | " 61 | \n",
1153 | " 85 | \n",
1154 | " 0 | \n",
1155 | " 5 | \n",
1156 | "
\n",
1157 | " \n",
1158 | " | 614.jpg | \n",
1159 | " 0 | \n",
1160 | " 0 | \n",
1161 | " 0 | \n",
1162 | " 0 | \n",
1163 | " 0 | \n",
1164 | " 0 | \n",
1165 | "
\n",
1166 | " \n",
1167 | " | 184.jpg | \n",
1168 | " 0 | \n",
1169 | " 0 | \n",
1170 | " 0 | \n",
1171 | " 0 | \n",
1172 | " 0 | \n",
1173 | " 0 | \n",
1174 | "
\n",
1175 | " \n",
1176 | " | 905.jpg | \n",
1177 | " 0 | \n",
1178 | " 0 | \n",
1179 | " 0 | \n",
1180 | " 0 | \n",
1181 | " 0 | \n",
1182 | " 0 | \n",
1183 | "
\n",
1184 | " \n",
1185 | " | 97.jpg | \n",
1186 | " 0 | \n",
1187 | " 0 | \n",
1188 | " 27 | \n",
1189 | " 11 | \n",
1190 | " 0 | \n",
1191 | " 0 | \n",
1192 | "
\n",
1193 | " \n",
1194 | " | 882.jpg | \n",
1195 | " 0 | \n",
1196 | " 0 | \n",
1197 | " 0 | \n",
1198 | " 0 | \n",
1199 | " 0 | \n",
1200 | " 0 | \n",
1201 | "
\n",
1202 | " \n",
1203 | " | 776.jpg | \n",
1204 | " 8 | \n",
1205 | " 2 | \n",
1206 | " 51 | \n",
1207 | " 2 | \n",
1208 | " 29 | \n",
1209 | " 6 | \n",
1210 | "
\n",
1211 | " \n",
1212 | " | 899.jpg | \n",
1213 | " 2 | \n",
1214 | " 2 | \n",
1215 | " 13 | \n",
1216 | " 3 | \n",
1217 | " 0 | \n",
1218 | " 0 | \n",
1219 | "
\n",
1220 | " \n",
1221 | " | 344.jpg | \n",
1222 | " 0 | \n",
1223 | " 0 | \n",
1224 | " 0 | \n",
1225 | " 0 | \n",
1226 | " 0 | \n",
1227 | " 0 | \n",
1228 | "
\n",
1229 | " \n",
1230 | " | 473.jpg | \n",
1231 | " 2 | \n",
1232 | " 0 | \n",
1233 | " 25 | \n",
1234 | " 60 | \n",
1235 | " 0 | \n",
1236 | " 1 | \n",
1237 | "
\n",
1238 | " \n",
1239 | " | 510.jpg | \n",
1240 | " 5 | \n",
1241 | " 1 | \n",
1242 | " 6 | \n",
1243 | " 0 | \n",
1244 | " 0 | \n",
1245 | " 0 | \n",
1246 | "
\n",
1247 | " \n",
1248 | " | 234.jpg | \n",
1249 | " 0 | \n",
1250 | " 0 | \n",
1251 | " 0 | \n",
1252 | " 0 | \n",
1253 | " 0 | \n",
1254 | " 0 | \n",
1255 | "
\n",
1256 | " \n",
1257 | " | 291.jpg | \n",
1258 | " 43 | \n",
1259 | " 20 | \n",
1260 | " 299 | \n",
1261 | " 24 | \n",
1262 | " 237 | \n",
1263 | " 11 | \n",
1264 | "
\n",
1265 | " \n",
1266 | " | 331.jpg | \n",
1267 | " 6 | \n",
1268 | " 18 | \n",
1269 | " 67 | \n",
1270 | " 0 | \n",
1271 | " 0 | \n",
1272 | " 5 | \n",
1273 | "
\n",
1274 | " \n",
1275 | " | 433.jpg | \n",
1276 | " 0 | \n",
1277 | " 3 | \n",
1278 | " 9 | \n",
1279 | " 43 | \n",
1280 | " 0 | \n",
1281 | " 3 | \n",
1282 | "
\n",
1283 | " \n",
1284 | " | 712.jpg | \n",
1285 | " 0 | \n",
1286 | " 0 | \n",
1287 | " 0 | \n",
1288 | " 0 | \n",
1289 | " 0 | \n",
1290 | " 0 | \n",
1291 | "
\n",
1292 | " \n",
1293 | " | 741.jpg | \n",
1294 | " 34 | \n",
1295 | " 5 | \n",
1296 | " 231 | \n",
1297 | " 38 | \n",
1298 | " 176 | \n",
1299 | " 1 | \n",
1300 | "
\n",
1301 | " \n",
1302 | " | 767.jpg | \n",
1303 | " 0 | \n",
1304 | " 0 | \n",
1305 | " 0 | \n",
1306 | " 0 | \n",
1307 | " 0 | \n",
1308 | " 0 | \n",
1309 | "
\n",
1310 | " \n",
1311 | " | 912.jpg | \n",
1312 | " 30 | \n",
1313 | " 2 | \n",
1314 | " 251 | \n",
1315 | " 45 | \n",
1316 | " 205 | \n",
1317 | " 432 | \n",
1318 | "
\n",
1319 | " \n",
1320 | "
\n",
1321 | "
\n",
1354 | "
\n",
1355 | " \n",
1356 | " \n",
1357 | " | \n",
1358 | " train_id | \n",
1359 | " adult_males | \n",
1360 | " subadult_males | \n",
1361 | " adult_females | \n",
1362 | " juveniles | \n",
1363 | " pups | \n",
1364 | "
\n",
1365 | " \n",
1366 | " \n",
1367 | " \n",
1368 | " | 531 | \n",
1369 | " 531 | \n",
1370 | " 0 | \n",
1371 | " 9 | \n",
1372 | " 5 | \n",
1373 | " 6 | \n",
1374 | " 0 | \n",
1375 | "
\n",
1376 | " \n",
1377 | " | 946 | \n",
1378 | " 946 | \n",
1379 | " 3 | \n",
1380 | " 1 | \n",
1381 | " 34 | \n",
1382 | " 33 | \n",
1383 | " 0 | \n",
1384 | "
\n",
1385 | " \n",
1386 | " | 34 | \n",
1387 | " 34 | \n",
1388 | " 4 | \n",
1389 | " 1 | \n",
1390 | " 127 | \n",
1391 | " 237 | \n",
1392 | " 0 | \n",
1393 | "
\n",
1394 | " \n",
1395 | " | 30 | \n",
1396 | " 30 | \n",
1397 | " 2 | \n",
1398 | " 0 | \n",
1399 | " 1 | \n",
1400 | " 3 | \n",
1401 | " 0 | \n",
1402 | "
\n",
1403 | " \n",
1404 | " | 290 | \n",
1405 | " 290 | \n",
1406 | " 2 | \n",
1407 | " 0 | \n",
1408 | " 0 | \n",
1409 | " 0 | \n",
1410 | " 0 | \n",
1411 | "
\n",
1412 | " \n",
1413 | " | 406 | \n",
1414 | " 406 | \n",
1415 | " 10 | \n",
1416 | " 12 | \n",
1417 | " 141 | \n",
1418 | " 47 | \n",
1419 | " 3 | \n",
1420 | "
\n",
1421 | " \n",
1422 | " | 380 | \n",
1423 | " 380 | \n",
1424 | " 2 | \n",
1425 | " 0 | \n",
1426 | " 83 | \n",
1427 | " 31 | \n",
1428 | " 0 | \n",
1429 | "
\n",
1430 | " \n",
1431 | " | 913 | \n",
1432 | " 913 | \n",
1433 | " 1 | \n",
1434 | " 0 | \n",
1435 | " 20 | \n",
1436 | " 33 | \n",
1437 | " 0 | \n",
1438 | "
\n",
1439 | " \n",
1440 | " | 621 | \n",
1441 | " 621 | \n",
1442 | " 15 | \n",
1443 | " 12 | \n",
1444 | " 53 | \n",
1445 | " 36 | \n",
1446 | " 30 | \n",
1447 | "
\n",
1448 | " \n",
1449 | " | 811 | \n",
1450 | " 811 | \n",
1451 | " 7 | \n",
1452 | " 5 | \n",
1453 | " 12 | \n",
1454 | " 0 | \n",
1455 | " 0 | \n",
1456 | "
\n",
1457 | " \n",
1458 | " | 7 | \n",
1459 | " 7 | \n",
1460 | " 6 | \n",
1461 | " 2 | \n",
1462 | " 33 | \n",
1463 | " 16 | \n",
1464 | " 3 | \n",
1465 | "
\n",
1466 | " \n",
1467 | " | 421 | \n",
1468 | " 421 | \n",
1469 | " 1 | \n",
1470 | " 3 | \n",
1471 | " 0 | \n",
1472 | " 0 | \n",
1473 | " 0 | \n",
1474 | "
\n",
1475 | " \n",
1476 | " | 292 | \n",
1477 | " 292 | \n",
1478 | " 2 | \n",
1479 | " 3 | \n",
1480 | " 23 | \n",
1481 | " 15 | \n",
1482 | " 1 | \n",
1483 | "
\n",
1484 | " \n",
1485 | " | 66 | \n",
1486 | " 66 | \n",
1487 | " 0 | \n",
1488 | " 0 | \n",
1489 | " 0 | \n",
1490 | " 0 | \n",
1491 | " 0 | \n",
1492 | "
\n",
1493 | " \n",
1494 | " | 593 | \n",
1495 | " 593 | \n",
1496 | " 1 | \n",
1497 | " 2 | \n",
1498 | " 32 | \n",
1499 | " 5 | \n",
1500 | " 0 | \n",
1501 | "
\n",
1502 | " \n",
1503 | " | 490 | \n",
1504 | " 490 | \n",
1505 | " 6 | \n",
1506 | " 5 | \n",
1507 | " 14 | \n",
1508 | " 4 | \n",
1509 | " 0 | \n",
1510 | "
\n",
1511 | " \n",
1512 | " | 909 | \n",
1513 | " 909 | \n",
1514 | " 5 | \n",
1515 | " 4 | \n",
1516 | " 30 | \n",
1517 | " 9 | \n",
1518 | " 0 | \n",
1519 | "
\n",
1520 | " \n",
1521 | " | 800 | \n",
1522 | " 800 | \n",
1523 | " 1 | \n",
1524 | " 1 | \n",
1525 | " 0 | \n",
1526 | " 0 | \n",
1527 | " 0 | \n",
1528 | "
\n",
1529 | " \n",
1530 | " | 215 | \n",
1531 | " 215 | \n",
1532 | " 0 | \n",
1533 | " 4 | \n",
1534 | " 0 | \n",
1535 | " 0 | \n",
1536 | " 0 | \n",
1537 | "
\n",
1538 | " \n",
1539 | " | 426 | \n",
1540 | " 426 | \n",
1541 | " 2 | \n",
1542 | " 5 | \n",
1543 | " 6 | \n",
1544 | " 32 | \n",
1545 | " 5 | \n",
1546 | "
\n",
1547 | " \n",
1548 | " | 475 | \n",
1549 | " 475 | \n",
1550 | " 3 | \n",
1551 | " 6 | \n",
1552 | " 61 | \n",
1553 | " 85 | \n",
1554 | " 0 | \n",
1555 | "
\n",
1556 | " \n",
1557 | " | 614 | \n",
1558 | " 614 | \n",
1559 | " 1 | \n",
1560 | " 0 | \n",
1561 | " 5 | \n",
1562 | " 10 | \n",
1563 | " 0 | \n",
1564 | "
\n",
1565 | " \n",
1566 | " | 184 | \n",
1567 | " 184 | \n",
1568 | " 6 | \n",
1569 | " 16 | \n",
1570 | " 19 | \n",
1571 | " 7 | \n",
1572 | " 0 | \n",
1573 | "
\n",
1574 | " \n",
1575 | " | 905 | \n",
1576 | " 905 | \n",
1577 | " 3 | \n",
1578 | " 2 | \n",
1579 | " 6 | \n",
1580 | " 0 | \n",
1581 | " 0 | \n",
1582 | "
\n",
1583 | " \n",
1584 | " | 97 | \n",
1585 | " 97 | \n",
1586 | " 0 | \n",
1587 | " 0 | \n",
1588 | " 19 | \n",
1589 | " 11 | \n",
1590 | " 0 | \n",
1591 | "
\n",
1592 | " \n",
1593 | " | 882 | \n",
1594 | " 882 | \n",
1595 | " 2 | \n",
1596 | " 4 | \n",
1597 | " 6 | \n",
1598 | " 16 | \n",
1599 | " 0 | \n",
1600 | "
\n",
1601 | " \n",
1602 | " | 776 | \n",
1603 | " 776 | \n",
1604 | " 3 | \n",
1605 | " 2 | \n",
1606 | " 25 | \n",
1607 | " 2 | \n",
1608 | " 29 | \n",
1609 | "
\n",
1610 | " \n",
1611 | " | 899 | \n",
1612 | " 899 | \n",
1613 | " 0 | \n",
1614 | " 3 | \n",
1615 | " 0 | \n",
1616 | " 2 | \n",
1617 | " 0 | \n",
1618 | "
\n",
1619 | " \n",
1620 | " | 344 | \n",
1621 | " 344 | \n",
1622 | " 2 | \n",
1623 | " 0 | \n",
1624 | " 18 | \n",
1625 | " 10 | \n",
1626 | " 0 | \n",
1627 | "
\n",
1628 | " \n",
1629 | " | 473 | \n",
1630 | " 473 | \n",
1631 | " 2 | \n",
1632 | " 0 | \n",
1633 | " 19 | \n",
1634 | " 57 | \n",
1635 | " 0 | \n",
1636 | "
\n",
1637 | " \n",
1638 | " | 510 | \n",
1639 | " 510 | \n",
1640 | " 2 | \n",
1641 | " 1 | \n",
1642 | " 0 | \n",
1643 | " 0 | \n",
1644 | " 0 | \n",
1645 | "
\n",
1646 | " \n",
1647 | " | 234 | \n",
1648 | " 234 | \n",
1649 | " 0 | \n",
1650 | " 2 | \n",
1651 | " 46 | \n",
1652 | " 25 | \n",
1653 | " 0 | \n",
1654 | "
\n",
1655 | " \n",
1656 | " | 291 | \n",
1657 | " 291 | \n",
1658 | " 39 | \n",
1659 | " 20 | \n",
1660 | " 275 | \n",
1661 | " 23 | \n",
1662 | " 237 | \n",
1663 | "
\n",
1664 | " \n",
1665 | " | 331 | \n",
1666 | " 331 | \n",
1667 | " 6 | \n",
1668 | " 18 | \n",
1669 | " 10 | \n",
1670 | " 0 | \n",
1671 | " 0 | \n",
1672 | "
\n",
1673 | " \n",
1674 | " | 433 | \n",
1675 | " 433 | \n",
1676 | " 0 | \n",
1677 | " 3 | \n",
1678 | " 5 | \n",
1679 | " 41 | \n",
1680 | " 0 | \n",
1681 | "
\n",
1682 | " \n",
1683 | " | 712 | \n",
1684 | " 712 | \n",
1685 | " 3 | \n",
1686 | " 0 | \n",
1687 | " 52 | \n",
1688 | " 64 | \n",
1689 | " 0 | \n",
1690 | "
\n",
1691 | " \n",
1692 | " | 741 | \n",
1693 | " 741 | \n",
1694 | " 32 | \n",
1695 | " 5 | \n",
1696 | " 228 | \n",
1697 | " 38 | \n",
1698 | " 176 | \n",
1699 | "
\n",
1700 | " \n",
1701 | " | 767 | \n",
1702 | " 767 | \n",
1703 | " 6 | \n",
1704 | " 7 | \n",
1705 | " 53 | \n",
1706 | " 14 | \n",
1707 | " 0 | \n",
1708 | "
\n",
1709 | " \n",
1710 | " | 912 | \n",
1711 | " 912 | \n",
1712 | " 30 | \n",
1713 | " 2 | \n",
1714 | " 244 | \n",
1715 | " 13 | \n",
1716 | " 205 | \n",
1717 | "
\n",
1718 | " \n",
1719 | "
\n",
1720 | "
\n",
1724 | "
\n",
1725 | " \n",
1726 | " \n",
1727 | " | \n",
1728 | " train_id | \n",
1729 | " adult_males | \n",
1730 | " subadult_males | \n",
1731 | " adult_females | \n",
1732 | " juveniles | \n",
1733 | " pups | \n",
1734 | "
\n",
1735 | " \n",
1736 | " \n",
1737 | " \n",
1738 | " | 531 | \n",
1739 | " 531 | \n",
1740 | " 0 | \n",
1741 | " 9 | \n",
1742 | " 5 | \n",
1743 | " 6 | \n",
1744 | " 0 | \n",
1745 | "
\n",
1746 | " \n",
1747 | " | 946 | \n",
1748 | " 946 | \n",
1749 | " 3 | \n",
1750 | " 1 | \n",
1751 | " 34 | \n",
1752 | " 33 | \n",
1753 | " 0 | \n",
1754 | "
\n",
1755 | " \n",
1756 | " | 34 | \n",
1757 | " 34 | \n",
1758 | " 4 | \n",
1759 | " 1 | \n",
1760 | " 127 | \n",
1761 | " 237 | \n",
1762 | " 0 | \n",
1763 | "
\n",
1764 | " \n",
1765 | " | 30 | \n",
1766 | " 30 | \n",
1767 | " 2 | \n",
1768 | " 0 | \n",
1769 | " 1 | \n",
1770 | " 3 | \n",
1771 | " 0 | \n",
1772 | "
\n",
1773 | " \n",
1774 | " | 290 | \n",
1775 | " 290 | \n",
1776 | " 2 | \n",
1777 | " 0 | \n",
1778 | " 0 | \n",
1779 | " 0 | \n",
1780 | " 0 | \n",
1781 | "
\n",
1782 | " \n",
1783 | " | 406 | \n",
1784 | " 406 | \n",
1785 | " 10 | \n",
1786 | " 12 | \n",
1787 | " 141 | \n",
1788 | " 47 | \n",
1789 | " 3 | \n",
1790 | "
\n",
1791 | " \n",
1792 | " | 380 | \n",
1793 | " 380 | \n",
1794 | " 2 | \n",
1795 | " 0 | \n",
1796 | " 83 | \n",
1797 | " 31 | \n",
1798 | " 0 | \n",
1799 | "
\n",
1800 | " \n",
1801 | " | 913 | \n",
1802 | " 913 | \n",
1803 | " 1 | \n",
1804 | " 0 | \n",
1805 | " 20 | \n",
1806 | " 33 | \n",
1807 | " 0 | \n",
1808 | "
\n",
1809 | " \n",
1810 | " | 621 | \n",
1811 | " 621 | \n",
1812 | " 15 | \n",
1813 | " 12 | \n",
1814 | " 53 | \n",
1815 | " 36 | \n",
1816 | " 30 | \n",
1817 | "
\n",
1818 | " \n",
1819 | " | 811 | \n",
1820 | " 811 | \n",
1821 | " 7 | \n",
1822 | " 5 | \n",
1823 | " 12 | \n",
1824 | " 0 | \n",
1825 | " 0 | \n",
1826 | "
\n",
1827 | " \n",
1828 | " | 7 | \n",
1829 | " 7 | \n",
1830 | " 6 | \n",
1831 | " 2 | \n",
1832 | " 33 | \n",
1833 | " 16 | \n",
1834 | " 3 | \n",
1835 | "
\n",
1836 | " \n",
1837 | " | 421 | \n",
1838 | " 421 | \n",
1839 | " 1 | \n",
1840 | " 3 | \n",
1841 | " 0 | \n",
1842 | " 0 | \n",
1843 | " 0 | \n",
1844 | "
\n",
1845 | " \n",
1846 | " | 292 | \n",
1847 | " 292 | \n",
1848 | " 2 | \n",
1849 | " 3 | \n",
1850 | " 23 | \n",
1851 | " 15 | \n",
1852 | " 1 | \n",
1853 | "
\n",
1854 | " \n",
1855 | " | 66 | \n",
1856 | " 66 | \n",
1857 | " 0 | \n",
1858 | " 0 | \n",
1859 | " 0 | \n",
1860 | " 0 | \n",
1861 | " 0 | \n",
1862 | "
\n",
1863 | " \n",
1864 | " | 593 | \n",
1865 | " 593 | \n",
1866 | " 1 | \n",
1867 | " 2 | \n",
1868 | " 32 | \n",
1869 | " 5 | \n",
1870 | " 0 | \n",
1871 | "
\n",
1872 | " \n",
1873 | " | 490 | \n",
1874 | " 490 | \n",
1875 | " 6 | \n",
1876 | " 5 | \n",
1877 | " 14 | \n",
1878 | " 4 | \n",
1879 | " 0 | \n",
1880 | "
\n",
1881 | " \n",
1882 | " | 909 | \n",
1883 | " 909 | \n",
1884 | " 5 | \n",
1885 | " 4 | \n",
1886 | " 30 | \n",
1887 | " 9 | \n",
1888 | " 0 | \n",
1889 | "
\n",
1890 | " \n",
1891 | " | 800 | \n",
1892 | " 800 | \n",
1893 | " 1 | \n",
1894 | " 1 | \n",
1895 | " 0 | \n",
1896 | " 0 | \n",
1897 | " 0 | \n",
1898 | "
\n",
1899 | " \n",
1900 | " | 215 | \n",
1901 | " 215 | \n",
1902 | " 0 | \n",
1903 | " 4 | \n",
1904 | " 0 | \n",
1905 | " 0 | \n",
1906 | " 0 | \n",
1907 | "
\n",
1908 | " \n",
1909 | " | 426 | \n",
1910 | " 426 | \n",
1911 | " 2 | \n",
1912 | " 5 | \n",
1913 | " 6 | \n",
1914 | " 32 | \n",
1915 | " 5 | \n",
1916 | "
\n",
1917 | " \n",
1918 | " | 475 | \n",
1919 | " 475 | \n",
1920 | " 3 | \n",
1921 | " 6 | \n",
1922 | " 61 | \n",
1923 | " 85 | \n",
1924 | " 0 | \n",
1925 | "
\n",
1926 | " \n",
1927 | " | 614 | \n",
1928 | " 614 | \n",
1929 | " 1 | \n",
1930 | " 0 | \n",
1931 | " 5 | \n",
1932 | " 10 | \n",
1933 | " 0 | \n",
1934 | "
\n",
1935 | " \n",
1936 | " | 184 | \n",
1937 | " 184 | \n",
1938 | " 6 | \n",
1939 | " 16 | \n",
1940 | " 19 | \n",
1941 | " 7 | \n",
1942 | " 0 | \n",
1943 | "
\n",
1944 | " \n",
1945 | " | 905 | \n",
1946 | " 905 | \n",
1947 | " 3 | \n",
1948 | " 2 | \n",
1949 | " 6 | \n",
1950 | " 0 | \n",
1951 | " 0 | \n",
1952 | "
\n",
1953 | " \n",
1954 | " | 97 | \n",
1955 | " 97 | \n",
1956 | " 0 | \n",
1957 | " 0 | \n",
1958 | " 19 | \n",
1959 | " 11 | \n",
1960 | " 0 | \n",
1961 | "
\n",
1962 | " \n",
1963 | " | 882 | \n",
1964 | " 882 | \n",
1965 | " 2 | \n",
1966 | " 4 | \n",
1967 | " 6 | \n",
1968 | " 16 | \n",
1969 | " 0 | \n",
1970 | "
\n",
1971 | " \n",
1972 | " | 776 | \n",
1973 | " 776 | \n",
1974 | " 3 | \n",
1975 | " 2 | \n",
1976 | " 25 | \n",
1977 | " 2 | \n",
1978 | " 29 | \n",
1979 | "
\n",
1980 | " \n",
1981 | " | 899 | \n",
1982 | " 899 | \n",
1983 | " 0 | \n",
1984 | " 3 | \n",
1985 | " 0 | \n",
1986 | " 2 | \n",
1987 | " 0 | \n",
1988 | "
\n",
1989 | " \n",
1990 | " | 344 | \n",
1991 | " 344 | \n",
1992 | " 2 | \n",
1993 | " 0 | \n",
1994 | " 18 | \n",
1995 | " 10 | \n",
1996 | " 0 | \n",
1997 | "
\n",
1998 | " \n",
1999 | " | 473 | \n",
2000 | " 473 | \n",
2001 | " 2 | \n",
2002 | " 0 | \n",
2003 | " 19 | \n",
2004 | " 57 | \n",
2005 | " 0 | \n",
2006 | "
\n",
2007 | " \n",
2008 | " | 510 | \n",
2009 | " 510 | \n",
2010 | " 2 | \n",
2011 | " 1 | \n",
2012 | " 0 | \n",
2013 | " 0 | \n",
2014 | " 0 | \n",
2015 | "
\n",
2016 | " \n",
2017 | " | 234 | \n",
2018 | " 234 | \n",
2019 | " 0 | \n",
2020 | " 2 | \n",
2021 | " 46 | \n",
2022 | " 25 | \n",
2023 | " 0 | \n",
2024 | "
\n",
2025 | " \n",
2026 | " | 291 | \n",
2027 | " 291 | \n",
2028 | " 39 | \n",
2029 | " 20 | \n",
2030 | " 275 | \n",
2031 | " 23 | \n",
2032 | " 237 | \n",
2033 | "
\n",
2034 | " \n",
2035 | " | 331 | \n",
2036 | " 331 | \n",
2037 | " 6 | \n",
2038 | " 18 | \n",
2039 | " 10 | \n",
2040 | " 0 | \n",
2041 | " 0 | \n",
2042 | "
\n",
2043 | " \n",
2044 | " | 433 | \n",
2045 | " 433 | \n",
2046 | " 0 | \n",
2047 | " 3 | \n",
2048 | " 5 | \n",
2049 | " 41 | \n",
2050 | " 0 | \n",
2051 | "
\n",
2052 | " \n",
2053 | " | 712 | \n",
2054 | " 712 | \n",
2055 | " 3 | \n",
2056 | " 0 | \n",
2057 | " 52 | \n",
2058 | " 64 | \n",
2059 | " 0 | \n",
2060 | "
\n",
2061 | " \n",
2062 | " | 741 | \n",
2063 | " 741 | \n",
2064 | " 32 | \n",
2065 | " 5 | \n",
2066 | " 228 | \n",
2067 | " 38 | \n",
2068 | " 176 | \n",
2069 | "
\n",
2070 | " \n",
2071 | " | 767 | \n",
2072 | " 767 | \n",
2073 | " 6 | \n",
2074 | " 7 | \n",
2075 | " 53 | \n",
2076 | " 14 | \n",
2077 | " 0 | \n",
2078 | "
\n",
2079 | " \n",
2080 | " | 912 | \n",
2081 | " 912 | \n",
2082 | " 30 | \n",
2083 | " 2 | \n",
2084 | " 244 | \n",
2085 | " 13 | \n",
2086 | " 205 | \n",
2087 | "
\n",
2088 | " \n",
2089 | "
\n",
2090 | "