├── .gitignore ├── LICENSE ├── README.md ├── asset └── arch.png ├── data └── OLIWER │ └── split.json ├── dataset.py ├── divide.py ├── evaluate.py ├── model.py ├── model_utils.py ├── preprocess.py ├── requirements.txt ├── test.py ├── utils.py └── weights └── model.pth /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # C extensions 7 | *.so 8 | 9 | # Distribution / packaging 10 | .Python 11 | build/ 12 | develop-eggs/ 13 | dist/ 14 | downloads/ 15 | eggs/ 16 | .eggs/ 17 | lib/ 18 | lib64/ 19 | parts/ 20 | sdist/ 21 | var/ 22 | wheels/ 23 | share/python-wheels/ 24 | *.egg-info/ 25 | .installed.cfg 26 | *.egg 27 | MANIFEST 28 | 29 | # PyInstaller 30 | # Usually these files are written by a python script from a template 31 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 32 | *.manifest 33 | *.spec 34 | 35 | # Installer logs 36 | pip-log.txt 37 | pip-delete-this-directory.txt 38 | 39 | # Unit test / coverage reports 40 | htmlcov/ 41 | .tox/ 42 | .nox/ 43 | .coverage 44 | .coverage.* 45 | .cache 46 | nosetests.xml 47 | coverage.xml 48 | *.cover 49 | *.py,cover 50 | .hypothesis/ 51 | .pytest_cache/ 52 | cover/ 53 | 54 | # Translations 55 | *.mo 56 | *.pot 57 | 58 | # Django stuff: 59 | *.log 60 | local_settings.py 61 | db.sqlite3 62 | db.sqlite3-journal 63 | 64 | # Flask stuff: 65 | instance/ 66 | .webassets-cache 67 | 68 | # Scrapy stuff: 69 | .scrapy 70 | 71 | # Sphinx documentation 72 | docs/_build/ 73 | 74 | # PyBuilder 75 | .pybuilder/ 76 | target/ 77 | 78 | # Jupyter Notebook 79 | .ipynb_checkpoints 80 | 81 | # IPython 82 | profile_default/ 83 | ipython_config.py 84 | 85 | # pyenv 86 | # For a library or package, you might want to ignore these files since the code is 87 | # intended to run in multiple environments; otherwise, check them in: 88 | # .python-version 89 | 90 | # pipenv 91 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. 92 | # However, in case of collaboration, if having platform-specific dependencies or dependencies 93 | # having no cross-platform support, pipenv may install dependencies that don't work, or not 94 | # install all needed dependencies. 95 | #Pipfile.lock 96 | 97 | # poetry 98 | # Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control. 99 | # This is especially recommended for binary packages to ensure reproducibility, and is more 100 | # commonly ignored for libraries. 101 | # https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control 102 | #poetry.lock 103 | 104 | # pdm 105 | # Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control. 106 | #pdm.lock 107 | # pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it 108 | # in version control. 109 | # https://pdm.fming.dev/latest/usage/project/#working-with-version-control 110 | .pdm.toml 111 | .pdm-python 112 | .pdm-build/ 113 | 114 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm 115 | __pypackages__/ 116 | 117 | # Celery stuff 118 | celerybeat-schedule 119 | celerybeat.pid 120 | 121 | # SageMath parsed files 122 | *.sage.py 123 | 124 | # Environments 125 | .env 126 | .venv 127 | env/ 128 | venv/ 129 | ENV/ 130 | env.bak/ 131 | venv.bak/ 132 | 133 | # Spyder project settings 134 | .spyderproject 135 | .spyproject 136 | 137 | # Rope project settings 138 | .ropeproject 139 | 140 | # mkdocs documentation 141 | /site 142 | 143 | # mypy 144 | .mypy_cache/ 145 | .dmypy.json 146 | dmypy.json 147 | 148 | # Pyre type checker 149 | .pyre/ 150 | 151 | # pytype static type analyzer 152 | .pytype/ 153 | 154 | # Cython debug symbols 155 | cython_debug/ 156 | 157 | # PyCharm 158 | # JetBrains specific template is maintained in a separate JetBrains.gitignore that can 159 | # be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore 160 | # and can be added to the global gitignore or merged into this file. For a more nuclear 161 | # option (not recommended) you can uncomment the following to ignore the entire idea folder. 162 | #.idea/ 163 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | 635 | Copyright (C) 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | Copyright (C) 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | #
:dolphin:DOLPHIN 2 | 3 |
4 | SCUT DLVC Lab 5 | Static Badge 6 | Static Badge 7 | Static Badge 8 | Static Badge 9 |

10 | 11 | Online Writer Retrieval with Chinese Handwritten Phrases: A Synergistic Temporal-Frequency Representation Learning Approach 12 | 13 | IEEE Transactions of Information Forensics and Security (TIFS), 2024 14 | 15 | :star:Official code of the DOLPHIN model and the release of the OLIWER dataset. 16 |
17 | 18 | ##
:ocean:Introduction
19 | 20 | DOLPHIN is an online writer retrieval model, designed to retrieve all online handwriting samples of a specific writer. It synergizes temporal and frequency learning to extract discriminative feature representations for online handwriting. 21 | 22 | ![DOLPHIN’s architecture](asset/arch.png) 23 | 24 |
The model architecture of DOLPIHN
25 | 26 | ##
:earth_asia:Environment
27 | 28 | ```bash 29 | git clone https://github.com/SCUT-DLVCLab/DOLPHIN.git 30 | cd DOLPHIN 31 | conda create -n dolphin python=3.8.16 32 | conda activate dolphin 33 | pip install -r requirements.txt 34 | ``` 35 | 36 | ##
:hammer_and_pick:Data Preparation
37 | 38 | Download the three subsets: CASIA-OLHWDB2, DCOH-E, and SCUT-COUCH2009 using the following links: 39 | 40 | - [Baidu Cloud](https://pan.baidu.com/s/1Op917v5IM7OushQ_xPNLSg?pwd=oler) 41 | - [Google Drive](https://drive.google.com/drive/folders/1W-R78wLSJXDhK998c_zIAEFxtPE10AX4?usp=sharing) 42 | 43 | Unzip the .zip archives using the following commands: 44 | 45 | ```bash 46 | unzip OLHWDB2.zip -d . 47 | unzip DCOH-E.zip -d . 48 | unzip COUCH09.zip -d . 49 | ``` 50 | 51 | The directory should look like this: 52 | 53 | ``` 54 | data-raw 55 | ├── COUCH09 56 | │ ├── 001 57 | │ └── ... 58 | ├── DCOH-E 59 | │ ├── dcoh-e313 60 | │ └── ... 61 | └── OLHWDB2 62 | ├── 001 63 | └── ... 64 | ``` 65 | 66 | Then run `preprocess.py` for data preprocessing: 67 | 68 | ```bash 69 | python preprocess.py --dataset olhwdb2 70 | python preprocess.py --dataset dcohe 71 | python preprocess.py --dataset couch 72 | ``` 73 | 74 | The preprocessed data will be saved at the `data` folder. 75 | 76 | Then run the `divide.py` to merge the three subsets into the **OLIWER** dataset and divide the data into `training` and `testing` parts. 77 | 78 | ```bash 79 | python divide.py --divide 80 | python divide.py --extract 81 | ``` 82 | 83 | Now the data should be all preprocessed. The final data directory should look like: 84 | 85 | ```bash 86 | data 87 | ├── COUCH09 88 | │ └── COUCH09.pkl 89 | ├── DCOH-E 90 | │ └── DCOH-E.pkl 91 | ├── OLHWDB2 92 | │ └── OLHWDB2.pkl 93 | └── OLIWER 94 | ├── split.json 95 | ├── test.pkl 96 | ├── test-tf.pkl 97 | ├── train.pkl 98 | └── train-tf.pkl 99 | ``` 100 | 101 | ##
:rocket:Test
102 | 103 | ``` 104 | python test.py --weights weights/model.pth 105 | ``` 106 | 107 | ##
:bookmark_tabs:Citation
108 | 109 | ```bibtex 110 | @ARTICLE{dolphin2024zhang, 111 | author={Zhang, Peirong and Jin, Lianwen}, 112 | journal={IEEE Transactions on Information Forensics and Security (TIFS)}, 113 | title={{Online Writer Retrieval With Chinese Handwritten Phrases: A Synergistic Temporal-Frequency Representation Learning Approach}}, 114 | year={2024}, 115 | volume={19}, 116 | number={}, 117 | pages={10387-10399} 118 | } 119 | ``` 120 | 121 | ##
:phone:Cotact
122 | 123 | Peirong Zhang: eeprzhang@mail.scut.edu.cn 124 | 125 | ##
:palm_tree:Copyright
126 | 127 | Copyright 2024-2025, Deep Learning and Vision Computing (DLVC) Lab, South China China University of Technology. [http://www.dlvc-lab.net](http://www.dlvc-lab.net/). 128 | 129 | -------------------------------------------------------------------------------- /asset/arch.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/SCUT-DLVCLab/DOLPHIN/fc90f848ef11cbdcc2f89fabf3a9ed68343641d0/asset/arch.png -------------------------------------------------------------------------------- /data/OLIWER/split.json: -------------------------------------------------------------------------------- 1 | { 2 | "train_writers": [ 3 | "dcoh-e416", 4 | "dcoh154", 5 | "1006", 6 | "120", 7 | "dcoh-e439", 8 | "106", 9 | "couch87", 10 | "740", 11 | "couch33", 12 | "639", 13 | "821", 14 | "804", 15 | "717", 16 | "687", 17 | "995", 18 | "dcoh61", 19 | "dcoh-e373", 20 | "912", 21 | "142", 22 | "790", 23 | "dcoh-e413", 24 | "419", 25 | "couch143", 26 | "12", 27 | "565", 28 | "dcoh13", 29 | "couch51", 30 | "567", 31 | "925", 32 | "dcoh264", 33 | "dcoh122", 34 | "754", 35 | "dcoh162", 36 | "dcoh223", 37 | "381", 38 | "dcoh-e392", 39 | "225", 40 | "753", 41 | "383", 42 | "dcoh-e313", 43 | "492", 44 | "dcoh62", 45 | "dcoh180", 46 | "dcoh276", 47 | "dcoh-e490", 48 | "dcoh-e421", 49 | "dcoh-e314", 50 | "749", 51 | "dcoh75", 52 | "dcoh-e508", 53 | "386", 54 | "683", 55 | "dcoh267", 56 | "382", 57 | "dcoh15", 58 | "712", 59 | "129", 60 | "dcoh151", 61 | "468", 62 | "775", 63 | "161", 64 | "931", 65 | "53", 66 | "dcoh-e411", 67 | "312", 68 | "dcoh-e519", 69 | "dcoh41", 70 | "448", 71 | "dcoh261", 72 | "17", 73 | "250", 74 | "893", 75 | "487", 76 | "242", 77 | "couch99", 78 | "961", 79 | "264", 80 | "dcoh92", 81 | "couch130", 82 | "455", 83 | "969", 84 | "dcoh-e456", 85 | "dcoh4", 86 | "43", 87 | "761", 88 | "883", 89 | "couch109", 90 | "dcoh-e484", 91 | "dcoh-e426", 92 | "733", 93 | "739", 94 | "dcoh-e533", 95 | "322", 96 | "dcoh-e328", 97 | "147", 98 | "dcoh150", 99 | "dcoh144", 100 | "168", 101 | "633", 102 | "dcoh-e345", 103 | "dcoh212", 104 | "dcoh269", 105 | "dcoh89", 106 | "couch20", 107 | "dcoh157", 108 | "335", 109 | "328", 110 | "807", 111 | "8", 112 | "826", 113 | "couch25", 114 | "dcoh298", 115 | "couch42", 116 | "773", 117 | "couch6", 118 | "65", 119 | "182", 120 | "couch21", 121 | "dcoh54", 122 | "295", 123 | "547", 124 | "30", 125 | "817", 126 | "688", 127 | "couch84", 128 | "dcoh-e385", 129 | "938", 130 | "606", 131 | "618", 132 | "373", 133 | "couch137", 134 | "dcoh-e381", 135 | "263", 136 | "305", 137 | "851", 138 | "138", 139 | "934", 140 | "811", 141 | "425", 142 | "36", 143 | "dcoh26", 144 | "dcoh48", 145 | "662", 146 | "233", 147 | "843", 148 | "877", 149 | "couch119", 150 | "80", 151 | "couch128", 152 | "329", 153 | "dcoh128", 154 | "357", 155 | "154", 156 | "764", 157 | "18", 158 | "dcoh-e403", 159 | "couch121", 160 | "couch27", 161 | "dcoh-e344", 162 | "dcoh-e468", 163 | "454", 164 | "806", 165 | "couch44", 166 | "906", 167 | "dcoh124", 168 | "589", 169 | "137", 170 | "dcoh178", 171 | "dcoh-e509", 172 | "963", 173 | "dcoh224", 174 | "dcoh-e459", 175 | "510", 176 | "dcoh-e316", 177 | "521", 178 | "443", 179 | "317", 180 | "473", 181 | "dcoh192", 182 | "413", 183 | "277", 184 | "729", 185 | "702", 186 | "813", 187 | "dcoh-e476", 188 | "824", 189 | "dcoh198", 190 | "715", 191 | "dcoh-e470", 192 | "672", 193 | "848", 194 | "couch60", 195 | "398", 196 | "759", 197 | "701", 198 | "967", 199 | "771", 200 | "370", 201 | "812", 202 | "dcoh-e474", 203 | "dcoh103", 204 | "943", 205 | "dcoh241", 206 | "1004", 207 | "couch58", 208 | "964", 209 | "209", 210 | "780", 211 | "347", 212 | "309", 213 | "132", 214 | "623", 215 | "805", 216 | "837", 217 | "881", 218 | "dcoh170", 219 | "dcoh60", 220 | "dcoh116", 221 | "713", 222 | "853", 223 | "735", 224 | "dcoh280", 225 | "dcoh219", 226 | "757", 227 | "153", 228 | "dcoh77", 229 | "353", 230 | "469", 231 | "808", 232 | "couch115", 233 | "1007", 234 | "285", 235 | "409", 236 | "362", 237 | "dcoh215", 238 | "dcoh-e354", 239 | "couch77", 240 | "689", 241 | "612", 242 | "81", 243 | "dcoh-e551", 244 | "252", 245 | "236", 246 | "614", 247 | "couch72", 248 | "600", 249 | "648", 250 | "872", 251 | "couch10", 252 | "dcoh272", 253 | "304", 254 | "dcoh-e451", 255 | "dcoh-e529", 256 | "527", 257 | "dcoh-e431", 258 | "dcoh-e471", 259 | "dcoh210", 260 | "41", 261 | "112", 262 | "675", 263 | "dcoh93", 264 | "485", 265 | "835", 266 | "581", 267 | "couch11", 268 | "dcoh-e318", 269 | "couch117", 270 | "dcoh200", 271 | "dcoh-e493", 272 | "282", 273 | "dcoh195", 274 | "202", 275 | "913", 276 | "couch7", 277 | "dcoh-e361", 278 | "993", 279 | "dcoh-e383", 280 | "96", 281 | "666", 282 | "couch118", 283 | "519", 284 | "722", 285 | "dcoh-e418", 286 | "dcoh-e343", 287 | "dcoh-e379", 288 | "dcoh46", 289 | "dcoh-e356", 290 | "725", 291 | "56", 292 | "165", 293 | "dcoh168", 294 | "64", 295 | "dcoh31", 296 | "324", 297 | "dcoh134", 298 | "694", 299 | "196", 300 | "couch136", 301 | "163", 302 | "763", 303 | "couch22", 304 | "798", 305 | "dcoh-e353", 306 | "49", 307 | "dcoh-e472", 308 | "dcoh250", 309 | "dcoh-e547", 310 | "16", 311 | "958", 312 | "dcoh-e401", 313 | "387", 314 | "dcoh-e412", 315 | "800", 316 | "940", 317 | "dcoh-e505", 318 | "92", 319 | "125", 320 | "dcoh57", 321 | "1012", 322 | "1008", 323 | "936", 324 | "593", 325 | "219", 326 | "844", 327 | "dcoh-e410", 328 | "888", 329 | "dcoh-e522", 330 | "dcoh-e521", 331 | "514", 332 | "653", 333 | "74", 334 | "228", 335 | "211", 336 | "355", 337 | "998", 338 | "569", 339 | "224", 340 | "dcoh-e402", 341 | "334", 342 | "248", 343 | "dcoh174", 344 | "dcoh-e534", 345 | "571", 346 | "352", 347 | "116", 348 | "150", 349 | "couch135", 350 | "449", 351 | "680", 352 | "95", 353 | "776", 354 | "928", 355 | "767", 356 | "868", 357 | "699", 358 | "542", 359 | "dcoh160", 360 | "54", 361 | "947", 362 | "307", 363 | "369", 364 | "couch75", 365 | "dcoh209", 366 | "1002", 367 | "992", 368 | "dcoh-e420", 369 | "707", 370 | "399", 371 | "dcoh-e495", 372 | "656", 373 | "dcoh294", 374 | "102", 375 | "230", 376 | "dcoh105", 377 | "751", 378 | "718", 379 | "dcoh101", 380 | "340", 381 | "774", 382 | "119", 383 | "260", 384 | "439", 385 | "dcoh211", 386 | "38", 387 | "396", 388 | "dcoh-e452", 389 | "dcoh0", 390 | "couch122", 391 | "458", 392 | "dcoh113", 393 | "dcoh2", 394 | "dcoh-e545", 395 | "dcoh-e473", 396 | "dcoh177", 397 | "578", 398 | "dcoh-e475", 399 | "736", 400 | "661", 401 | "dcoh-e329", 402 | "dcoh-e520", 403 | "dcoh71", 404 | "156", 405 | "395", 406 | "dcoh19", 407 | "802", 408 | "407", 409 | "507", 410 | "dcoh-e461", 411 | "870", 412 | "177", 413 | "dcoh238", 414 | "204", 415 | "dcoh63", 416 | "954", 417 | "dcoh32", 418 | "270", 419 | "919", 420 | "dcoh-e478", 421 | "dcoh-e400", 422 | "couch35", 423 | "dcoh106", 424 | "dcoh8", 425 | "684", 426 | "dcoh88", 427 | "555", 428 | "971", 429 | "784", 430 | "dcoh-e555", 431 | "568", 432 | "247", 433 | "358", 434 | "dcoh129", 435 | "397", 436 | "889", 437 | "105", 438 | "619", 439 | "dcoh259", 440 | "986", 441 | "couch62", 442 | "dcoh-e393", 443 | "148", 444 | "830", 445 | "538", 446 | "166", 447 | "dcoh74", 448 | "39", 449 | "couch73", 450 | "842", 451 | "143", 452 | "dcoh-e325", 453 | "couch19", 454 | "218", 455 | "1018", 456 | "627", 457 | "311", 458 | "652", 459 | "768", 460 | "dcoh-e527", 461 | "447", 462 | "620", 463 | "187", 464 | "831", 465 | "676", 466 | "428", 467 | "dcoh9", 468 | "couch68", 469 | "977", 470 | "28", 471 | "dcoh234", 472 | "dcoh68", 473 | "354", 474 | "85", 475 | "86", 476 | "dcoh-e457", 477 | "dcoh312", 478 | "dcoh278", 479 | "985", 480 | "dcoh-e434", 481 | "275", 482 | "880", 483 | "306", 484 | "dcoh45", 485 | "dcoh-e557", 486 | "couch64", 487 | "908", 488 | "113", 489 | "378", 490 | "77", 491 | "302", 492 | "dcoh-e357", 493 | "320", 494 | "803", 495 | "dcoh254", 496 | "dcoh85", 497 | "535", 498 | "couch52", 499 | "dcoh-e364", 500 | "dcoh187", 501 | "642", 502 | "dcoh208", 503 | "dcoh226", 504 | "640", 505 | "511", 506 | "dcoh-e358", 507 | "390", 508 | "886", 509 | "dcoh-e396", 510 | "dcoh266", 511 | "dcoh-e566", 512 | "159", 513 | "0", 514 | "dcoh303", 515 | "dcoh-e567", 516 | "232", 517 | "dcoh-e513", 518 | "341", 519 | "601", 520 | "852", 521 | "dcoh-e377", 522 | "414", 523 | "dcoh127", 524 | "dcoh-e319", 525 | "dcoh-e481", 526 | "185", 527 | "175", 528 | "292", 529 | "269", 530 | "786", 531 | "622", 532 | "229", 533 | "207", 534 | "332", 535 | "253", 536 | "109", 537 | "dcoh283", 538 | "636", 539 | "couch34", 540 | "549", 541 | "dcoh-e515", 542 | "200", 543 | "dcoh-e324", 544 | "couch45", 545 | "dcoh-e347", 546 | "110", 547 | "dcoh-e532", 548 | "337", 549 | "789", 550 | "69", 551 | "dcoh175", 552 | "dcoh-e376", 553 | "dcoh-e510", 554 | "dcoh167", 555 | "635", 556 | "488", 557 | "647", 558 | "dcoh-e453", 559 | "dcoh43", 560 | "900", 561 | "couch90", 562 | "58", 563 | "22", 564 | "dcoh-e525", 565 | "dcoh-e367", 566 | "dcoh-e550", 567 | "dcoh227", 568 | "couch96", 569 | "84", 570 | "couch120", 571 | "couch100", 572 | "674", 573 | "dcoh-e553", 574 | "42", 575 | "dcoh-e561", 576 | "610", 577 | "288", 578 | "dcoh40", 579 | "dcoh-e443", 580 | "158", 581 | "dcoh-e503", 582 | "604", 583 | "7", 584 | "520", 585 | "62", 586 | "dcoh-e560", 587 | "dcoh-e339", 588 | "dcoh251", 589 | "408", 590 | "748", 591 | "dcoh-e494", 592 | "573", 593 | "465", 594 | "489", 595 | "195", 596 | "400", 597 | "646", 598 | "dcoh110", 599 | "617", 600 | "couch85", 601 | "345", 602 | "couch53", 603 | "559", 604 | "dcoh-e386", 605 | "couch106", 606 | "couch67", 607 | "11", 608 | "83", 609 | "couch39", 610 | "dcoh291", 611 | "dcoh18", 612 | "dcoh76", 613 | "dcoh118", 614 | "couch138", 615 | "664", 616 | "75", 617 | "4", 618 | "391", 619 | "dcoh17", 620 | "176", 621 | "609", 622 | "536", 623 | "279", 624 | "dcoh-e415", 625 | "227", 626 | "dcoh142", 627 | "719", 628 | "dcoh-e432", 629 | "dcoh-e430", 630 | "dcoh309", 631 | "dcoh-e466", 632 | "dcoh70", 633 | "88", 634 | "dcoh-e563", 635 | "278", 636 | "dcoh-e408", 637 | "dcoh-e445", 638 | "486", 639 | "dcoh277", 640 | "dcoh-e317", 641 | "couch112", 642 | "660", 643 | "894", 644 | "dcoh130", 645 | "215", 646 | "290", 647 | "44", 648 | "dcoh-e487", 649 | "603", 650 | "dcoh237", 651 | "404", 652 | "dcoh28", 653 | "371", 654 | "539", 655 | "453", 656 | "couch108", 657 | "360", 658 | "186", 659 | "990", 660 | "dcoh-e336", 661 | "162", 662 | "494", 663 | "dcoh-e539", 664 | "460", 665 | "720", 666 | "711", 667 | "331", 668 | "673", 669 | "dcoh27", 670 | "500", 671 | "452", 672 | "dcoh279", 673 | "866", 674 | "1", 675 | "823", 676 | "dcoh107", 677 | "couch82", 678 | "173", 679 | "dcoh58", 680 | "24", 681 | "dcoh-e365", 682 | "338", 683 | "628", 684 | "497", 685 | "645", 686 | "couch61", 687 | "737", 688 | "671", 689 | "778", 690 | "474", 691 | "dcoh136", 692 | "dcoh-e341", 693 | "dcoh146", 694 | "210", 695 | "dcoh-e501", 696 | "350", 697 | "467", 698 | "dcoh23", 699 | "dcoh240", 700 | "couch16", 701 | "70", 702 | "1016", 703 | "420", 704 | "dcoh100", 705 | "dcoh-e372", 706 | "dcoh184", 707 | "957", 708 | "579", 709 | "dcoh-e424", 710 | "734", 711 | "898", 712 | "dcoh64", 713 | "393", 714 | "dcoh53", 715 | "199", 716 | "651", 717 | "300", 718 | "dcoh176", 719 | "couch14", 720 | "dcoh189", 721 | "dcoh-e334", 722 | "dcoh-e435", 723 | "863", 724 | "dcoh-e331", 725 | "68", 726 | "133", 727 | "dcoh164", 728 | "107", 729 | "couch57", 730 | "273", 731 | "dcoh232", 732 | "couch101", 733 | "459", 734 | "dcoh-e399", 735 | "942", 736 | "502", 737 | "dcoh-e359", 738 | "283", 739 | "384", 740 | "97", 741 | "655", 742 | "19", 743 | "819", 744 | "827", 745 | "418", 746 | "dcoh143", 747 | "205", 748 | "dcoh161", 749 | "1014", 750 | "52", 751 | "543", 752 | "dcoh256", 753 | "424", 754 | "dcoh-e467", 755 | "203", 756 | "214", 757 | "dcoh72", 758 | "dcoh290", 759 | "289", 760 | "dcoh222", 761 | "191", 762 | "couch134", 763 | "couch29", 764 | "239", 765 | "dcoh-e500", 766 | "dcoh307", 767 | "dcoh-e371", 768 | "478", 769 | "couch94", 770 | "10", 771 | "dcoh36", 772 | "745", 773 | "910", 774 | "832", 775 | "dcoh293", 776 | "couch133", 777 | "dcoh235", 778 | "dcoh218", 779 | "dcoh-e465", 780 | "825", 781 | "dcoh6", 782 | "343", 783 | "585", 784 | "dcoh42", 785 | "93", 786 | "dcoh172", 787 | "couch17", 788 | "518", 789 | "911", 790 | "226", 791 | "920", 792 | "692", 793 | "557", 794 | "602", 795 | "607", 796 | "845", 797 | "693", 798 | "788", 799 | "644", 800 | "643", 801 | "586", 802 | "couch56", 803 | "584", 804 | "141", 805 | "couch98", 806 | "374", 807 | "659", 808 | "couch69", 809 | "801", 810 | "136", 811 | "dcoh-e544", 812 | "611", 813 | "57", 814 | "dcoh-e323", 815 | "904", 816 | "479", 817 | "dcoh-e389", 818 | "941", 819 | "933", 820 | "dcoh29", 821 | "854", 822 | "1003", 823 | "190", 824 | "705", 825 | "dcoh102", 826 | "couch1", 827 | "dcoh49", 828 | "dcoh-e441", 829 | "280", 830 | "598", 831 | "244", 832 | "dcoh-e338", 833 | "412", 834 | "55", 835 | "dcoh51", 836 | "172", 837 | "couch50", 838 | "dcoh112", 839 | "dcoh-e541", 840 | "365", 841 | "dcoh119", 842 | "couch31", 843 | "327", 844 | "dcoh-e488", 845 | "94", 846 | "dcoh271", 847 | "dcoh59", 848 | "dcoh-e395", 849 | "dcoh-e460", 850 | "dcoh56", 851 | "dcoh-e498", 852 | "couch43", 853 | "dcoh5", 854 | "342", 855 | "dcoh-e531", 856 | "dcoh260", 857 | "291", 858 | "545", 859 | "dcoh73", 860 | "couch5", 861 | "couch113", 862 | "dcoh201", 863 | "dcoh228", 864 | "192", 865 | "dcoh138", 866 | "388", 867 | "dcoh22", 868 | "563", 869 | "dcoh247", 870 | "895", 871 | "833", 872 | "476", 873 | "756", 874 | "456", 875 | "dcoh-e427", 876 | "921", 877 | "dcoh38", 878 | "445", 879 | "dcoh1", 880 | "29", 881 | "548", 882 | "975", 883 | "1017", 884 | "dcoh308", 885 | "40", 886 | "208", 887 | "couch144", 888 | "978", 889 | "dcoh-e448", 890 | "742", 891 | "795", 892 | "180", 893 | "dcoh263", 894 | "902", 895 | "268", 896 | "71", 897 | "164", 898 | "dcoh243", 899 | "couch8", 900 | "266", 901 | "670", 902 | "698", 903 | "couch9", 904 | "336", 905 | "828", 906 | "35", 907 | "686", 908 | "989", 909 | "446", 910 | "178", 911 | "149", 912 | "dcoh193", 913 | "743", 914 | "dcoh-e446", 915 | "dcoh194", 916 | "882", 917 | "530", 918 | "430", 919 | "dcoh39", 920 | "727", 921 | "couch81", 922 | "dcoh83", 923 | "980", 924 | "389", 925 | "couch66", 926 | "988", 927 | "dcoh246", 928 | "544", 929 | "dcoh248", 930 | "654", 931 | "220", 932 | "dcoh-e320", 933 | "dcoh86", 934 | "968", 935 | "couch114", 936 | "368", 937 | "344", 938 | "dcoh-e538", 939 | "dcoh253", 940 | "364", 941 | "1005", 942 | "dcoh206", 943 | "738", 944 | "892", 945 | "dcoh-e499", 946 | "907", 947 | "dcoh-e437", 948 | "dcoh242", 949 | "426", 950 | "couch15", 951 | "323", 952 | "758", 953 | "couch86", 954 | "3", 955 | "867", 956 | "dcoh-e485", 957 | "552", 958 | "695", 959 | "dcoh-e438", 960 | "615", 961 | "929", 962 | "433", 963 | "972", 964 | "dcoh299", 965 | "dcoh109", 966 | "376", 967 | "532", 968 | "937", 969 | "dcoh-e483", 970 | "126", 971 | "dcoh126", 972 | "dcoh179", 973 | "dcoh-e407", 974 | "dcoh11", 975 | "dcoh166", 976 | "213", 977 | "528", 978 | "dcoh-e506", 979 | "dcoh155", 980 | "couch129", 981 | "dcoh310", 982 | "435", 983 | "783", 984 | "483", 985 | "375", 986 | "678", 987 | "682", 988 | "657", 989 | "dcoh-e504", 990 | "dcoh-e558", 991 | "484", 992 | "couch70", 993 | "dcoh295", 994 | "couch104", 995 | "436", 996 | "couch105", 997 | "dcoh-e440", 998 | "265", 999 | "couch132", 1000 | "couch0", 1001 | "170", 1002 | "dcoh-e480", 1003 | "529", 1004 | "dcoh-e479", 1005 | "couch59", 1006 | "710", 1007 | "709", 1008 | "dcoh-e349", 1009 | "dcoh95", 1010 | "dcoh-e335", 1011 | "dcoh-e414", 1012 | "198", 1013 | "308", 1014 | "157", 1015 | "131", 1016 | "700", 1017 | "couch83", 1018 | "98", 1019 | "dcoh233", 1020 | "377", 1021 | "27", 1022 | "951", 1023 | "924", 1024 | "235", 1025 | "791", 1026 | "dcoh245", 1027 | "dcoh-e326", 1028 | "dcoh52", 1029 | "dcoh287", 1030 | "974", 1031 | "859", 1032 | "dcoh-e464", 1033 | "dcoh205", 1034 | "708", 1035 | "505", 1036 | "dcoh-e375", 1037 | "490", 1038 | "379", 1039 | "dcoh33", 1040 | "108", 1041 | "184", 1042 | "dcoh90", 1043 | "189", 1044 | "dcoh7", 1045 | "dcoh-e535", 1046 | "506", 1047 | "366", 1048 | "222", 1049 | "582", 1050 | "810", 1051 | "566", 1052 | "553", 1053 | "349", 1054 | "couch142", 1055 | "822", 1056 | "267", 1057 | "471", 1058 | "dcoh220", 1059 | "286", 1060 | "792", 1061 | "82", 1062 | "72", 1063 | "860", 1064 | "couch78", 1065 | "781", 1066 | "271", 1067 | "dcoh173", 1068 | "79", 1069 | "356", 1070 | "dcoh156", 1071 | "dcoh203", 1072 | "481", 1073 | "dcoh-e536", 1074 | "34", 1075 | "dcoh-e450", 1076 | "525", 1077 | "dcoh-e442", 1078 | "dcoh152", 1079 | "dcoh-e423", 1080 | "dcoh111", 1081 | "dcoh133", 1082 | "256", 1083 | "495", 1084 | "561", 1085 | "427", 1086 | "dcoh-e542", 1087 | "442", 1088 | "930", 1089 | "984", 1090 | "76", 1091 | "67", 1092 | "457", 1093 | "223", 1094 | "1013", 1095 | "965", 1096 | "99", 1097 | "130", 1098 | "915", 1099 | "234", 1100 | "918", 1101 | "785", 1102 | "23", 1103 | "couch55", 1104 | "dcoh183", 1105 | "127", 1106 | "dcoh-e537", 1107 | "dcoh-e562", 1108 | "73", 1109 | "319", 1110 | "dcoh-e404", 1111 | "516", 1112 | "429", 1113 | "976", 1114 | "973", 1115 | "dcoh216", 1116 | "909", 1117 | "770", 1118 | "592", 1119 | "dcoh-e454", 1120 | "dcoh296", 1121 | "dcoh66", 1122 | "dcoh196", 1123 | "691", 1124 | "590", 1125 | "272", 1126 | "193", 1127 | "dcoh-e444", 1128 | "953", 1129 | "dcoh-e337", 1130 | "923", 1131 | "884", 1132 | "dcoh-e512", 1133 | "477", 1134 | "dcoh-e321", 1135 | "46", 1136 | "534", 1137 | "dcoh286", 1138 | "dcoh-e489", 1139 | "dcoh84", 1140 | "577", 1141 | "333", 1142 | "772", 1143 | "922", 1144 | "dcoh207", 1145 | "914", 1146 | "944", 1147 | "351", 1148 | "732", 1149 | "313", 1150 | "couch13", 1151 | "dcoh82", 1152 | "dcoh-e502", 1153 | "724", 1154 | "dcoh-e390", 1155 | "couch37", 1156 | "1000", 1157 | "297", 1158 | "779", 1159 | "905", 1160 | "171", 1161 | "dcoh-e482", 1162 | "385", 1163 | "450", 1164 | "dcoh-e394", 1165 | "183", 1166 | "dcoh10", 1167 | "51", 1168 | "501", 1169 | "dcoh-e564", 1170 | "dcoh21", 1171 | "583", 1172 | "596", 1173 | "dcoh-e322", 1174 | "dcoh-e548", 1175 | "361", 1176 | "750", 1177 | "87", 1178 | "405", 1179 | "515", 1180 | "31", 1181 | "849", 1182 | "dcoh96", 1183 | "dcoh273", 1184 | "dcoh121", 1185 | "310", 1186 | "dcoh139", 1187 | "901", 1188 | "dcoh-e405", 1189 | "217", 1190 | "dcoh-e497", 1191 | "917", 1192 | "dcoh301", 1193 | "996", 1194 | "15", 1195 | "726", 1196 | "dcoh188", 1197 | "441", 1198 | "194", 1199 | "1010", 1200 | "48", 1201 | "728", 1202 | "dcoh-e425", 1203 | "dcoh-e524", 1204 | "14", 1205 | "637", 1206 | "523", 1207 | "246", 1208 | "dcoh-e492", 1209 | "13", 1210 | "318", 1211 | "983", 1212 | "dcoh104", 1213 | "152", 1214 | "249", 1215 | "410", 1216 | "299", 1217 | "461", 1218 | "576", 1219 | "6", 1220 | "couch92", 1221 | "259", 1222 | "952", 1223 | "101", 1224 | "625", 1225 | "560", 1226 | "394", 1227 | "855", 1228 | "dcoh-e429", 1229 | "787", 1230 | "dcoh114", 1231 | "994", 1232 | "dcoh229", 1233 | "dcoh305", 1234 | "dcoh149", 1235 | "couch139", 1236 | "294", 1237 | "dcoh311", 1238 | "685", 1239 | "dcoh80", 1240 | "dcoh97", 1241 | "437", 1242 | "dcoh87", 1243 | "dcoh257", 1244 | "couch97", 1245 | "couch24", 1246 | "537", 1247 | "846", 1248 | "dcoh165", 1249 | "dcoh249", 1250 | "dcoh44", 1251 | "681", 1252 | "955", 1253 | "956", 1254 | "couch26", 1255 | "111", 1256 | "123", 1257 | "dcoh-e528", 1258 | "508", 1259 | "dcoh284", 1260 | "dcoh-e428", 1261 | "181", 1262 | "899", 1263 | "690", 1264 | "380", 1265 | "dcoh-e458", 1266 | "dcoh-e496", 1267 | "838", 1268 | "dcoh-e511", 1269 | "151", 1270 | "dcoh-e327", 1271 | "221", 1272 | "556", 1273 | "926", 1274 | "couch2", 1275 | "dcoh147", 1276 | "517", 1277 | "dcoh132", 1278 | "dcoh231", 1279 | "dcoh37", 1280 | "533", 1281 | "dcoh-e355", 1282 | "703", 1283 | "dcoh-e374", 1284 | "503", 1285 | "122", 1286 | "782", 1287 | "couch49", 1288 | "couch141", 1289 | "946", 1290 | "890", 1291 | "couch48", 1292 | "839", 1293 | "564", 1294 | "864", 1295 | "755", 1296 | "470", 1297 | "546", 1298 | "dcoh50", 1299 | "145", 1300 | "262", 1301 | "dcoh-e491", 1302 | "188", 1303 | "dcoh-e540", 1304 | "856", 1305 | "couch63", 1306 | "dcoh-e340", 1307 | "66", 1308 | "60", 1309 | "631", 1310 | "301", 1311 | "777", 1312 | "dcoh304", 1313 | "167", 1314 | "dcoh159", 1315 | "couch107", 1316 | "dcoh236", 1317 | "dcoh78", 1318 | "dcoh135", 1319 | "624", 1320 | "dcoh282", 1321 | "650", 1322 | "887", 1323 | "dcoh275", 1324 | "dcoh-e348", 1325 | "couch116", 1326 | "dcoh181", 1327 | "862", 1328 | "couch124", 1329 | "509", 1330 | "dcoh117", 1331 | "dcoh25", 1332 | "959", 1333 | "couch93", 1334 | "634", 1335 | "dcoh285", 1336 | "couch88", 1337 | "339", 1338 | "dcoh131", 1339 | "179", 1340 | "595", 1341 | "861", 1342 | "dcoh281", 1343 | "697", 1344 | "816", 1345 | "103", 1346 | "201", 1347 | "dcoh-e397", 1348 | "dcoh-e369", 1349 | "367", 1350 | "dcoh65", 1351 | "608", 1352 | "couch47", 1353 | "dcoh-e409", 1354 | "677", 1355 | "dcoh67", 1356 | "499", 1357 | "dcoh306", 1358 | "117", 1359 | "dcoh-e333", 1360 | "dcoh141", 1361 | "dcoh274", 1362 | "dcoh-e342", 1363 | "997", 1364 | "dcoh169", 1365 | "316", 1366 | "couch140", 1367 | "couch4", 1368 | "1009", 1369 | "dcoh-e315", 1370 | "dcoh47", 1371 | "dcoh-e417", 1372 | "couch65", 1373 | "981", 1374 | "61", 1375 | "124", 1376 | "421", 1377 | "dcoh288", 1378 | "1001", 1379 | "dcoh148", 1380 | "760", 1381 | "621", 1382 | "927", 1383 | "281", 1384 | "dcoh91", 1385 | "858", 1386 | "dcoh-e391" 1387 | ], 1388 | "test_writers": [ 1389 | "325", 1390 | "258", 1391 | "couch41", 1392 | "104", 1393 | "dcoh30", 1394 | "couch54", 1395 | "296", 1396 | "couch125", 1397 | "935", 1398 | "169", 1399 | "160", 1400 | "840", 1401 | "730", 1402 | "dcoh292", 1403 | "dcoh163", 1404 | "504", 1405 | "dcoh-e330", 1406 | "dcoh-e477", 1407 | "498", 1408 | "721", 1409 | "dcoh-e455", 1410 | "dcoh-e549", 1411 | "couch38", 1412 | "663", 1413 | "597", 1414 | "363", 1415 | "793", 1416 | "dcoh265", 1417 | "638", 1418 | "dcoh-e433", 1419 | "287", 1420 | "522", 1421 | "dcoh14", 1422 | "723", 1423 | "dcoh-e523", 1424 | "dcoh-e517", 1425 | "dcoh-e526", 1426 | "20", 1427 | "dcoh-e546", 1428 | "dcoh244", 1429 | "879", 1430 | "326", 1431 | "128", 1432 | "dcoh-e514", 1433 | "89", 1434 | "873", 1435 | "9", 1436 | "couch23", 1437 | "869", 1438 | "799", 1439 | "669", 1440 | "875", 1441 | "423", 1442 | "114", 1443 | "216", 1444 | "541", 1445 | "231", 1446 | "580", 1447 | "979", 1448 | "118", 1449 | "274", 1450 | "480", 1451 | "couch3", 1452 | "865", 1453 | "dcoh-e422", 1454 | "797", 1455 | "438", 1456 | "667", 1457 | "dcoh153", 1458 | "276", 1459 | "251", 1460 | "769", 1461 | "couch46", 1462 | "570", 1463 | "dcoh-e530", 1464 | "dcoh-e449", 1465 | "850", 1466 | "496", 1467 | "dcoh270", 1468 | "321", 1469 | "794", 1470 | "26", 1471 | "couch89", 1472 | "dcoh20", 1473 | "970", 1474 | "dcoh-e370", 1475 | "5", 1476 | "couch76", 1477 | "747", 1478 | "540", 1479 | "878", 1480 | "814", 1481 | "couch102", 1482 | "440", 1483 | "641", 1484 | "752", 1485 | "dcoh-e363", 1486 | "432", 1487 | "704", 1488 | "197", 1489 | "491", 1490 | "962", 1491 | "dcoh-e398", 1492 | "939", 1493 | "dcoh-e559", 1494 | "dcoh-e362", 1495 | "dcoh-e332", 1496 | "146", 1497 | "731", 1498 | "616", 1499 | "dcoh262", 1500 | "444", 1501 | "dcoh-e406", 1502 | "100", 1503 | "dcoh197", 1504 | "59", 1505 | "dcoh-e554", 1506 | "613", 1507 | "couch103", 1508 | "679", 1509 | "couch131", 1510 | "63", 1511 | "dcoh182", 1512 | "257", 1513 | "706", 1514 | "dcoh12", 1515 | "415", 1516 | "372", 1517 | "78", 1518 | "809", 1519 | "431", 1520 | "982", 1521 | "dcoh-e366", 1522 | "966", 1523 | "couch36", 1524 | "462", 1525 | "dcoh35", 1526 | "dcoh-e552", 1527 | "couch32", 1528 | "couch79", 1529 | "couch91", 1530 | "588", 1531 | "434", 1532 | "dcoh99", 1533 | "245", 1534 | "411", 1535 | "couch40", 1536 | "632", 1537 | "couch74", 1538 | "284", 1539 | "475", 1540 | "couch30", 1541 | "dcoh158", 1542 | "dcoh-e462", 1543 | "dcoh16", 1544 | "744", 1545 | "couch12", 1546 | "dcoh221", 1547 | "696", 1548 | "37", 1549 | "dcoh255", 1550 | "1015", 1551 | "885", 1552 | "212", 1553 | "847", 1554 | "dcoh185", 1555 | "dcoh69", 1556 | "626", 1557 | "dcoh81", 1558 | "couch18", 1559 | "dcoh98", 1560 | "21", 1561 | "dcoh258", 1562 | "dcoh268", 1563 | "422", 1564 | "dcoh302", 1565 | "dcoh252", 1566 | "144", 1567 | "714", 1568 | "298", 1569 | "665", 1570 | "820", 1571 | "couch80", 1572 | "575", 1573 | "315", 1574 | "dcoh-e382", 1575 | "466", 1576 | "dcoh145", 1577 | "359", 1578 | "558", 1579 | "916", 1580 | "134", 1581 | "dcoh-e518", 1582 | "554", 1583 | "couch28", 1584 | "562", 1585 | "241", 1586 | "dcoh94", 1587 | "330", 1588 | "746", 1589 | "dcoh204", 1590 | "dcoh125", 1591 | "950", 1592 | "couch126", 1593 | "174", 1594 | "dcoh217", 1595 | "572", 1596 | "dcoh-e507", 1597 | "402", 1598 | "348", 1599 | "591", 1600 | "303", 1601 | "551", 1602 | "829", 1603 | "dcoh-e368", 1604 | "dcoh34", 1605 | "524", 1606 | "91", 1607 | "dcoh230", 1608 | "dcoh171", 1609 | "716", 1610 | "dcoh289", 1611 | "897", 1612 | "couch71", 1613 | "dcoh-e436", 1614 | "416", 1615 | "dcoh-e378", 1616 | "574", 1617 | "dcoh225", 1618 | "dcoh191", 1619 | "493", 1620 | "629", 1621 | "dcoh-e380", 1622 | "couch123", 1623 | "dcoh140", 1624 | "741", 1625 | "dcoh-e516", 1626 | "33", 1627 | "135", 1628 | "949", 1629 | "658", 1630 | "932", 1631 | "630", 1632 | "987", 1633 | "dcoh-e447", 1634 | "dcoh-e556", 1635 | "dcoh190", 1636 | "531", 1637 | "605", 1638 | "dcoh213", 1639 | "90", 1640 | "818", 1641 | "47", 1642 | "dcoh120", 1643 | "dcoh55", 1644 | "couch95", 1645 | "243", 1646 | "406", 1647 | "dcoh115", 1648 | "dcoh-e351", 1649 | "796", 1650 | "dcoh-e360", 1651 | "dcoh-e384", 1652 | "dcoh108", 1653 | "891", 1654 | "762", 1655 | "115", 1656 | "dcoh202", 1657 | "874", 1658 | "834", 1659 | "451", 1660 | "261", 1661 | "2", 1662 | "513", 1663 | "871", 1664 | "140", 1665 | "32", 1666 | "991", 1667 | "206", 1668 | "dcoh199", 1669 | "472", 1670 | "255", 1671 | "401", 1672 | "dcoh123", 1673 | "dcoh297", 1674 | "841", 1675 | "254", 1676 | "240", 1677 | "948", 1678 | "50", 1679 | "45", 1680 | "594", 1681 | "945", 1682 | "587", 1683 | "464", 1684 | "550", 1685 | "766", 1686 | "dcoh-e419", 1687 | "dcoh137", 1688 | "couch110", 1689 | "121", 1690 | "960", 1691 | "dcoh-e387", 1692 | "dcoh300", 1693 | "482", 1694 | "dcoh239", 1695 | "dcoh-e486", 1696 | "25", 1697 | "dcoh214", 1698 | "765", 1699 | "903", 1700 | "512", 1701 | "649", 1702 | "dcoh-e346", 1703 | "417", 1704 | "896", 1705 | "403", 1706 | "dcoh-e543", 1707 | "dcoh-e469", 1708 | "857", 1709 | "526", 1710 | "dcoh-e463", 1711 | "155", 1712 | "463", 1713 | "dcoh24", 1714 | "815", 1715 | "293", 1716 | "314", 1717 | "couch111", 1718 | "999", 1719 | "couch127", 1720 | "392", 1721 | "836", 1722 | "dcoh-e352", 1723 | "dcoh-e565", 1724 | "dcoh-e350", 1725 | "668", 1726 | "139", 1727 | "238", 1728 | "1011", 1729 | "dcoh3", 1730 | "346", 1731 | "dcoh-e388", 1732 | "599", 1733 | "876", 1734 | "dcoh79", 1735 | "237" 1736 | ] 1737 | } -------------------------------------------------------------------------------- /dataset.py: -------------------------------------------------------------------------------- 1 | # -*- coding:utf-8 -*- 2 | 3 | import numpy as np 4 | from torch.utils.data import Dataset 5 | 6 | class Writing(Dataset): 7 | def __init__(self,handwriting_info:dict,transform=None,train=True): 8 | super().__init__() 9 | self.users = handwriting_info.keys() 10 | self.users_cnt = len(self.users) 11 | self.train = train 12 | self.features = [] 13 | self.user_labels = [] 14 | for i,k in enumerate(self.users): 15 | # extract_features(handwriting_info[k],self.features) 16 | self.features.extend(handwriting_info[k]) 17 | self.user_labels.extend([i] * len(handwriting_info[k])) 18 | assert len(self.user_labels) == len(self.features) 19 | self.features_cnt = len(self.features) 20 | self.feature_dims = np.shape(self.features[0])[1] # 就是时间函数的数量，这里是12个 21 | self.transform = transform 22 | 23 | def __len__(self): 24 | return self.features_cnt 25 | 26 | def __getitem__(self,idx): 27 | if self.train: 28 | feature = self.features[idx] 29 | if self.transform is not None: 30 | feature = self.transform(feature) 31 | else: 32 | feature = self.features[idx] 33 | return feature,len(feature),self.user_labels[idx] 34 | 35 | def collate_fn(batch:list): 36 | batch_size = len(batch) 37 | handwriting = [i[0] for i in batch] 38 | hw_len = np.array([i[1] for i in batch],dtype=np.float32) 39 | user_labels = np.array([i[2] for i in batch]) 40 | max_len = int(np.max(hw_len)) 41 | time_function_cnts = np.shape(handwriting[0])[1] 42 | handwriting_padded = np.zeros((batch_size,max_len,time_function_cnts),dtype=np.float32) 43 | for i,hw in enumerate(handwriting): 44 | handwriting_padded[i,:hw.shape[0]] = hw 45 | return handwriting_padded,hw_len,user_labels 46 | 47 | if __name__ == '__main__': 48 | ... -------------------------------------------------------------------------------- /divide.py: -------------------------------------------------------------------------------- 1 | # -*- coding:utf-8 -*- 2 | 3 | import os,pickle,json,argparse 4 | import numpy as np 5 | from utils import time_functions,clock 6 | 7 | def divide_data(src_root='./data',tgt_root='./data/OLIWER'): 8 | if not os.path.exists(f'{tgt_root}/OLIWER.pkl'): 9 | print('Merging data.') 10 | with open(f'{src_root}/OLHWDB2/OLHWDB2.pkl','rb') as f: 11 | olhwdb2 = pickle.load(f,encoding='iso-8859-1') 12 | with open(f'{src_root}/DCOH-E/DCOH-E.pkl','rb') as f: 13 | dcohe = pickle.load(f,encoding='iso-8859-1') 14 | with open(f'{src_root}/COUCH09/COUCH09.pkl','rb') as f: 15 | couch = pickle.load(f,encoding='iso-8859-1') 16 | 17 | olhwdb_cnt,dcohe_cnt,couch_cnt = 0,0,0 18 | olhwdb_writer_cnt,dcohe_writer_cnt,couch_writer_cnt = 0,0,0 19 | data = {} 20 | for k in olhwdb2: 21 | if len(olhwdb2[k]) <= 20: 22 | continue 23 | data[str(k)] = olhwdb2[k] 24 | olhwdb_cnt += len(olhwdb2[k]) 25 | olhwdb_writer_cnt += 1 26 | for k in couch: 27 | if len(couch[k]) <= 20: 28 | continue 29 | newk = f'couch{k}' 30 | data[newk] = couch[k] 31 | couch_cnt += len(couch[k]) 32 | 33 | couch_writer_cnt += 1 34 | for k in dcohe: 35 | if len(dcohe[k]) <= 20: 36 | continue 37 | data[k] = dcohe[k] 38 | dcohe_cnt += len(dcohe[k]) 39 | dcohe_writer_cnt += 1 40 | cnt = 0 41 | for k in data: 42 | cnt += len(data[k]) 43 | print('user:',len(data),'sample:',cnt) 44 | print('dcohe samples:',dcohe_cnt,dcohe_writer_cnt) 45 | print('olhwdb2 samples:',olhwdb_cnt,olhwdb_writer_cnt) 46 | print('couch samples:',couch_cnt,couch_writer_cnt) 47 | os.makedirs(tgt_root,exist_ok=True) 48 | with open(f'{tgt_root}/OLIWER.pkl','wb') as f: 49 | pickle.dump(data,f) 50 | else: 51 | print('Loading existing data.') 52 | with open(f'{tgt_root}/OLIWER.pkl','rb') as f: 53 | data = pickle.load(f,encoding='iso-8859-1') 54 | print('user:',len(data),'sample:',np.sum([len(data[k]) for k in data.keys()])) 55 | 56 | if os.path.exists(f'./{tgt_root}/split.json'): 57 | with open(f'{tgt_root}/split.json') as f: 58 | split = json.load(f) 59 | train_writers = split['train_writers'] 60 | test_writers = split['test_writers'] 61 | print('Loading existing splits.') 62 | else: 63 | train_num = int(0.8 * len(data.keys())) 64 | train_writers = np.random.choice(list(data.keys()),size=train_num,replace=False) 65 | test_writers = list(set(list(data.keys())) - set(train_writers)) 66 | split = {} 67 | split['train_writers'] = list(train_writers) 68 | split['test_writers'] = list(test_writers) 69 | with open(f'{tgt_root}/split.json','w',encoding='utf-8') as f: 70 | f.write(json.dumps(split,indent=4,ensure_ascii=False)) 71 | print('Generating new splits.') 72 | 73 | train,test = {},{} 74 | for k in train_writers: 75 | train[k] = data[k] 76 | for k in test_writers: 77 | test[k] = data[k] 78 | 79 | with open(f'{tgt_root}/train.pkl','wb') as f: 80 | pickle.dump(train,f) 81 | with open(f'{tgt_root}/test.pkl','wb') as f: 82 | pickle.dump(test,f) 83 | 84 | @clock 85 | def extract_and_store(src_root='./data/OLIWER/train.pkl',tgt_root='./data/OLIWER/train-tf.pkl'): 86 | with open(src_root,'rb') as f: 87 | handwriting_info = pickle.load(f,encoding='iso-8859-1') 88 | writing = {} 89 | for i,k in enumerate(handwriting_info.keys()): 90 | writing[k] = [] 91 | for each in handwriting_info[k]: 92 | writing[k].append(time_functions(each)) 93 | with open(tgt_root,'wb') as f: 94 | pickle.dump(writing,f) 95 | 96 | if __name__ == '__main__': 97 | parser = argparse.ArgumentParser() 98 | parser.add_argument('--divide',action='store_true') 99 | parser.add_argument('--extract',action='store_true') 100 | opt = parser.parse_args() 101 | if opt.divide: 102 | divide_data('./data','./data/OLIWER') 103 | if opt.extract: 104 | extract_and_store('./data/OLIWER/train.pkl','./data/OLIWER/train-tf.pkl') 105 | extract_and_store('./data/OLIWER/test.pkl','./data/OLIWER/test-tf.pkl') -------------------------------------------------------------------------------- /evaluate.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | 3 | import torch,os,time 4 | from torch import nn 5 | from scipy import io 6 | import numpy as np 7 | from model_utils import db_augmentation,average_query_expansion 8 | 9 | def evaluate(qf,ql,gf,gl): # q是query，g是gallery 10 | # q是一个，gallery是全部19732个 11 | # print(qf.shape,ql.shape,qc.shape,gf.shape,gl.shape,gc.shape) 12 | query = qf.view(-1,1) # (512,1) 13 | score = torch.mm(gf,query) # (19732,1) 14 | # score = nn.PairwiseDistance(p=2)(gf,qf.view(1,-1)) 15 | score = score.squeeze().cpu().numpy() 16 | 17 | idx = np.argsort(score)[::-1] 18 | # print(ql,gl[idx][:10],gl[idx][10:20],gl[idx][20:30]) 19 | query_idx = np.argwhere(gl == ql) # gallery里总有当前query的类别的，而且不止一张图是这个类 20 | positive_idx = query_idx 21 | metrics = compute_mAP(idx,positive_idx) 22 | return metrics # (ap,CMC) 23 | 24 | def compute_mAP(idx,positive_idx): 25 | ap = 0 26 | cmc = torch.zeros((len(idx))) 27 | if positive_idx.size == 0: 28 | cmc[0] = -1 29 | return ap,cmc 30 | len_pos = len(positive_idx) 31 | mask = np.in1d(idx,positive_idx) # 找到positive的index，也就是不同摄像头的同一个人 32 | rows_pos = np.argwhere(mask).flatten() 33 | # print(rows_pos,len_pos) 34 | cmc[rows_pos[0]:] = 1 # 赋值1的每个位置都不同，每个位置累加起来再除以总数就是Acc@1这些了 35 | # 注意这里有个:,是从第一个往后都赋值1，所以Rank@10会比Rank@1大 36 | for i in range(len_pos): # len_pos是不知道的，所以无所谓有多少个gallery样本 37 | precision = (i + 1) * 1. / (rows_pos[i] + 1) # 这就是每一个格子的precision 38 | if rows_pos[i] != 0: 39 | old_precision = i * 1.0 / rows_pos[i] 40 | else: 41 | old_precision = 1.0 42 | ap = ap + (old_precision + precision) / 2 # 不太理解为什么要old_precision然后除以2 43 | # ap = ap + precision 44 | ap = ap / len_pos 45 | 46 | return ap,cmc 47 | 48 | def compute_metrics(res,logger,dba,device,verbose=True): 49 | query_feature = res['query_feature'] 50 | query_label = res['query_label'] 51 | gallery_feature = res['gallery_feature'] 52 | gallery_label = res['gallery_label'] 53 | 54 | if dba: 55 | time_start = time.time() 56 | query_feature,gallery_feature = db_augmentation(query_feature,gallery_feature,10) 57 | query_feature,gallery_feature = average_query_expansion(query_feature,gallery_feature,5) 58 | query_feature = query_feature / np.linalg.norm(query_feature,axis=1,keepdims=True) 59 | gallery_feature = gallery_feature / np.linalg.norm(gallery_feature,axis=1,keepdims=True) 60 | logger.info(f'DBA & AQE time consuming: {time.time() - time_start:.4f}s') 61 | 62 | query_feature = torch.FloatTensor(query_feature).to(device) 63 | gallery_feature = torch.FloatTensor(gallery_feature).to(device) 64 | 65 | CMC = torch.zeros((len(gallery_label))) 66 | # aps = [] 67 | ap = 0. 68 | time_sum = 0. 69 | for i in range(len(query_label)): 70 | time_start = time.time() 71 | cur_ap,cur_CMC = evaluate(query_feature[i],query_label[i],gallery_feature,gallery_label) 72 | time_sum += (time.time() - time_start) 73 | if cur_CMC[0] == -1: continue 74 | CMC += cur_CMC 75 | ap += cur_ap 76 | # aps.append(cur_ap) 77 | # logger.info(f'evaluate time consuming: {time_sum:.4f}s') 78 | time_avg = time_sum / len(query_label) 79 | 80 | CMC /= len(query_label) 81 | ap /= len(query_label) 82 | if verbose: 83 | logger.info(f'[single query] Rank@1: {CMC[0] * 100.:.4f}% Rank@5: {CMC[4] * 100.:.4f}% Rank@10: {CMC[9] * 100.:.4f}%') 84 | logger.info(f'[single query] mAP: {ap * 100.:.4f}%') 85 | return time_avg,ap,CMC[0] * 100.,CMC[4] * 100.,CMC[9] * 100. 86 | 87 | if __name__ == '__main__': 88 | from dataset import Writing 89 | import pickle 90 | import matplotlib.pyplot as plt 91 | gallery_root = f'./data/OLER/gallery-tf-optxy2.pkl' 92 | with open(gallery_root,'rb') as f: 93 | handwriting_info = pickle.load(f,encoding='iso-8859-1') 94 | gallery_dataset = Writing(handwriting_info,train=False) 95 | _,aps = compute_metrics(None,False,'cpu') 96 | l = gallery_dataset.user_labels 97 | print(aps) 98 | k = [len(np.where(l == i)[0]) for i in np.sort(list(set(l)))] 99 | print(k) 100 | d = {i:(len(np.where(l == i)[0]),aps[i]) for i in np.sort(list(set(l)))} 101 | 102 | d1 = [(len(np.where(l == i)[0]),aps[i]) for i in np.sort(list(set(l)))] 103 | d1 = sorted(d1,key=lambda x:x[0]) 104 | # plt.hist(aps,len(aps)) 105 | # x_array = list(set(d.values())) 106 | # plt.bar(range(x_array),) 107 | aps = [each[1] for each in d1] 108 | idx = [each[0] for each in d1] 109 | idx1 = [idx[0],idx[len(idx) // 4],idx[len(idx) // 2],idx[len(idx) // 4 * 3],idx[-1]] 110 | idx2 = [0,len(idx) // 4,len(idx) // 2,len(idx) // 4 * 3,len(idx)] 111 | idx3 = [''] * len(idx) 112 | idx3[0] = idx[0] 113 | idx3[len(idx) // 4] = idx[len(idx) // 4] 114 | idx3[len(idx) // 2] = idx[len(idx) // 2] 115 | idx3[len(idx) // 4 * 3] = idx[len(idx) // 4 * 3] 116 | idx3[-1] = idx[-1] 117 | print(d1) 118 | plt.bar(range(len(aps)),aps) 119 | plt.xticks(range(len(idx)),idx3) 120 | plt.savefig('./kkk.png',dpi=500) 121 | # print(aps) 122 | # a = np.array([1,2,3,4,5,1,2,1,1]) 123 | # b = np.array([3,2]) 124 | # c = np.in1d(a,b) 125 | # print(c) -------------------------------------------------------------------------------- /model.py: -------------------------------------------------------------------------------- 1 | # -*- coding:utf-8 -*- 2 | 3 | import torch 4 | import torch.nn as nn 5 | import torch.nn.functional as F 6 | from model_utils import DepthwiseSeparableConv,CAIR,HFGA,ContextBlock,SelectivePool1d,get_len_mask 7 | 8 | class Head(nn.Module): 9 | def __init__(self,d_in,d_hidden,num_classes,bias=True): 10 | super().__init__() 11 | self.fc1 = nn.Linear(d_in,d_hidden,bias=bias) 12 | self.dropout = nn.Dropout(0.1) 13 | self.head = nn.Linear(d_hidden,num_classes,bias=bias) 14 | 15 | def forward(self,x): 16 | x = self.dropout(self.fc1(x)) 17 | y = self.head(x) 18 | return x,y 19 | 20 | class DOLPHIN(nn.Module): 21 | def __init__(self,d_in,num_classes): 22 | super().__init__() 23 | self.conv = DepthwiseSeparableConv(d_in,64,7,skip=False,se_rate=0.,drop_path_rate=0.) 24 | self.block0 = nn.Sequential( 25 | CAIR(64,96,1,kernel_size=5,stride=2,skip=False,se_rate=0.25,drop_path_rate=0.1), 26 | CAIR(96,128,3,kernel_size=3,stride=1,skip=False,se_rate=0.25,drop_path_rate=0.1) 27 | ) 28 | self.block1 = nn.Sequential( 29 | CAIR(128,160,1,kernel_size=5,stride=2,skip=False,se_rate=0.25,drop_path_rate=0.1), 30 | CAIR(160,192,3,kernel_size=3,stride=1,skip=False,se_rate=0.25,drop_path_rate=0.1) 31 | ) 32 | self.block2 = nn.Sequential( 33 | CAIR(192,224,1,kernel_size=5,stride=2,skip=False,se_rate=0.25,drop_path_rate=0.1), 34 | CAIR(224,256,3,kernel_size=3,stride=1,skip=False,se_rate=0.25,drop_path_rate=0.1), 35 | ) 36 | self.freq1 = HFGA(128) 37 | self.freq_proj1 = DepthwiseSeparableConv(128,192,stride=2,skip=False,se_rate=0.,drop_path_rate=0.) 38 | self.freq2 = HFGA(192) 39 | self.freq_proj2 = DepthwiseSeparableConv(192,256,stride=2,skip=False,se_rate=0.,drop_path_rate=0.) 40 | self.freq3 = HFGA(256) 41 | self.up = nn.Upsample(scale_factor=2) 42 | self.conv_proj1 = nn.Conv1d(192,256,1,padding=0) 43 | self.conv_proj2 = nn.Conv1d(128,256,1,padding=0) 44 | self.context1 = ContextBlock(256,256 // 8) 45 | self.context2 = ContextBlock(256,256 // 8) 46 | self.context3 = ContextBlock(256,256 // 8) 47 | self.context4 = ContextBlock(256,256 // 8) 48 | self.head = Head(384 * 2,384,num_classes,bias=False) 49 | self.sel_pool1 = SelectivePool1d(256,d_head=24,num_heads=16) 50 | self.sel_pool2 = SelectivePool1d(256,d_head=24,num_heads=16) 51 | self.sel_pool3 = SelectivePool1d(256,d_head=24,num_heads=16) 52 | self.freq_head = nn.Sequential( 53 | nn.Linear(384,384,bias=False), 54 | nn.Dropout(0.1), 55 | ) 56 | self.weights_init() 57 | 58 | def forward(self,x,feature_lens): 59 | x = x.transpose(1,2) 60 | x = self.conv(x) 61 | y0 = self.block0(x) # (n,128,l/2) 62 | freq1 = self.freq1(y0) 63 | 64 | y1 = y0 + freq1 65 | y1 = self.block1(y1) # (n,192,l/4) 66 | freq1 = self.freq_proj1(freq1) 67 | freq1 = freq1 + y1 68 | freq2 = self.freq2(freq1) 69 | 70 | y2 = y1 + freq2 71 | y2 = self.block2(y2) # (n,256,l/8) 72 | freq2 = self.freq_proj2(freq2) 73 | freq2 = freq2 + y2 74 | freq3 = self.freq3(freq2) 75 | 76 | y3 = self.context1(self.up(y2)[:,:,:y1.shape[2]]) + self.context2(self.conv_proj1(y1)) # (n,320,l/8) 77 | y4 = self.context3(self.up(y3)[:,:,:y0.shape[2]]) + self.context4(self.conv_proj2(y0)) # (n,320,l/4) 78 | y3 = F.selu(y3,inplace=True) 79 | y4 = F.selu(y4,inplace=True) 80 | 81 | feature_lens = torch.div(feature_lens + 1,2,rounding_mode='trunc') 82 | mask = get_len_mask(feature_lens) 83 | f1 = self.sel_pool1(y4,mask) 84 | mask = F.max_pool1d(mask,2,ceil_mode=True) 85 | f2 = self.sel_pool2(y3,mask) 86 | f3 = self.sel_pool3(freq3,None) 87 | y_vector = torch.cat([f1,f2],dim=1) 88 | y_vector,y_prob = self.head(y_vector) 89 | f3 = self.freq_head(f3) 90 | return y_vector,y_prob,f3 91 | 92 | def weights_init(self): 93 | for m in self.modules(): 94 | if isinstance(m,nn.Conv1d): 95 | nn.init.kaiming_normal_(m.weight,mode='fan_out',a=1) 96 | if m.bias is not None: 97 | nn.init.zeros_(m.bias) 98 | elif isinstance(m,nn.Linear): 99 | nn.init.kaiming_normal_(m.weight,a=1) 100 | if m.bias is not None: 101 | nn.init.zeros_(m.bias) -------------------------------------------------------------------------------- /model_utils.py: -------------------------------------------------------------------------------- 1 | # -*- coding:utf-8 -*- 2 | 3 | import torch 4 | import torch.nn as nn 5 | import torch.nn.functional as F 6 | import numpy as np 7 | from timm.models.layers import trunc_normal_ 8 | from pytorch_wavelets import DWT1D,IDWT1D 9 | 10 | class SelectivePool1d(nn.Module): 11 | def __init__(self,in_features,d_head,num_heads): 12 | super().__init__() 13 | self.keys = nn.Parameter(torch.Tensor(num_heads,d_head),requires_grad=True) 14 | self.W_q = nn.Conv1d(in_features,d_head * num_heads,kernel_size=1) 15 | self.norm = 1 / np.sqrt(d_head) 16 | self.d_head = d_head 17 | self.num_heads = num_heads 18 | self.weights_init() 19 | 20 | def weights_init(self): 21 | nn.init.orthogonal_(self.keys,gain=1) 22 | nn.init.kaiming_normal_(self.W_q.weight,a=1) 23 | nn.init.zeros_(self.W_q.bias) 24 | 25 | def orthogonal_norm(self): 26 | keys = F.normalize(self.keys,dim=1) 27 | corr = torch.mm(keys,keys.transpose(0,1)) 28 | return torch.sum(torch.triu(corr,1).abs_()) 29 | 30 | def forward(self,x,mask): 31 | N,_,L = x.shape # (N,C,L) 32 | q = v = self.W_q(x).transpose(1,2).view(N,L,self.num_heads,self.d_head) 33 | if mask is not None: 34 | mask = mask.to(x.device) 35 | attn = F.softmax(torch.sum(q * self.keys,dim=-1) * self.norm - (1. - mask).unsqueeze(2) * 1000,dim=1) 36 | # (N,L,num_heads) 37 | else: 38 | attn = F.softmax(torch.sum(q * self.keys,dim=-1) * self.norm,dim=1) 39 | y = torch.sum(v * attn.unsqueeze(3),dim=1).view(N,-1) # (N,d_head * num_heads) 40 | return y 41 | 42 | def get_len_mask(features_lens): # mask需要更新，因为长度不一样 43 | features_lens = features_lens 44 | batch_size = len(features_lens) 45 | max_len = torch.max(features_lens) 46 | mask = torch.zeros((batch_size,max_len),dtype=torch.float32) 47 | for i in range(batch_size): 48 | mask[i,0:features_lens[i]] = 1.0 49 | return mask 50 | 51 | class Swish(nn.Module): 52 | def forward(self,x): 53 | return x * torch.sigmoid(x) 54 | 55 | class SwishImpl(torch.autograd.Function): 56 | @staticmethod 57 | def forward(ctx,i): 58 | res = i * torch.sigmoid(i) 59 | ctx.save_for_backward(i) 60 | return res 61 | 62 | @staticmethod 63 | def backward(ctx,y_grad): 64 | i = ctx.saved_tensors[0] 65 | x_sigmoid = torch.sigmoid(i) 66 | return y_grad * (x_sigmoid * (1 + i * (1 - x_sigmoid))) 67 | 68 | class MemoryEfficientSwish(nn.Module): 69 | def forward(self,x): 70 | return SwishImpl.apply(x) 71 | 72 | class SEBlock2(nn.Module): # 通道不一样 73 | def __init__(self,d_in,d_hidden,act_layer=Swish): # Swish或SiLU 74 | super().__init__() 75 | self.fc = nn.Sequential( 76 | nn.AdaptiveAvgPool1d(1), 77 | nn.Conv1d(d_in,d_hidden,kernel_size=1,padding=0,stride=1), 78 | act_layer(), 79 | nn.Conv1d(d_hidden,d_in,kernel_size=1,padding=0,stride=1), 80 | nn.Sigmoid()) 81 | 82 | def forward(self,x): # x: (n,c,l) 83 | y = self.fc(x) 84 | return x * y.expand_as(x) 85 | 86 | def compute_similarity(query,gallery): 87 | query = query / np.linalg.norm(query,axis=1,keepdims=True) 88 | gallery = gallery / np.linalg.norm(gallery,axis=1,keepdims=True) 89 | return np.matmul(query,gallery.T) 90 | 91 | def db_augmentation(query,gallery,topk=10): 92 | # DBA: Database-side feature augmentation https://link.springer.com/article/10.1007/s11263-017-1016-8 93 | weights = np.logspace(0,-2.,topk + 1) 94 | 95 | # query augmentation 96 | similarity = compute_similarity(query,gallery) 97 | indices = np.argsort(-similarity,axis=1) 98 | topk_gallery = gallery[indices[:,:topk],:] 99 | query = np.tensordot(weights,np.concatenate([query[:,None],topk_gallery],axis=1),axes=(0,1)) 100 | 101 | # gallery augmentation 102 | similarity = compute_similarity(gallery,gallery) 103 | indices = np.argsort(-similarity,axis=1) 104 | topk_gallery = gallery[indices[:,:topk + 1],:] 105 | gallery = np.tensordot(weights,topk_gallery,axes=(0,1)) 106 | return query,gallery 107 | 108 | def average_query_expansion(query,gallery,topk=5): 109 | similarity = compute_similarity(query,gallery) 110 | indices = np.argsort(-similarity,axis=1) 111 | topk_gallery = np.mean(gallery[indices[:,:topk],:],axis=1) 112 | query = np.concatenate([query,topk_gallery],axis=1) 113 | 114 | similarity = compute_similarity(gallery,gallery) 115 | indices = np.argsort(-similarity,axis=1) 116 | topk_gallery = np.mean(gallery[indices[:,1:topk + 1],:],axis=1) 117 | gallery = np.concatenate([gallery,topk_gallery],axis=1) 118 | return query,gallery 119 | def __init__(self,d_feat,l): 120 | super().__init__() 121 | self.conv_depth = nn.Conv1d(d_feat,d_feat,kernel_size=3,padding=1,bias=False,groups=d_feat // 2) 122 | self.complex_weight = nn.Parameter(torch.randn(d_feat,l,2,dtype=torch.float32) * 0.02) 123 | trunc_normal_(self.complex_weight,std=.02) 124 | self.head = nn.Linear(d_feat,d_feat,bias=True) 125 | 126 | def forward(self,x): 127 | x1 = x[:,:,0::2] 128 | x2 = x[:,:,1::2] 129 | x1 = self.conv_depth(x1) 130 | _,_,l = x2.shape 131 | x2 = torch.fft.rfft(x2,dim=2,norm='ortho') 132 | weight = self.complex_weight 133 | if not weight.shape[1:2] == x2.shape[2:3]: 134 | weight = F.interpolate(weight.permute(2,0,1).unsqueeze(2),size=(1,x2.shape[2]),mode='bilinear',align_corners=True).squeeze().permute(1,2,0) 135 | weight = torch.view_as_complex(weight.contiguous()) 136 | x2 *= weight 137 | x2 = torch.fft.irfft(x2,n=l,dim=2,norm='ortho') 138 | y = x1 + x2 139 | y = self.head(y.transpose(1,2)).transpose(1,2) 140 | return y 141 | 142 | def channel_shuffle(x,groups): 143 | n,c,l = x.shape 144 | d_hidden = c // groups 145 | x = x.view(n,groups,d_hidden,l) 146 | x = x.transpose(1,2).contiguous() 147 | x = x.view(n,-1,l) 148 | return x 149 | 150 | class ShuffleBlock(nn.Module): # 这就是ShufflNetV2的简化实现，本来就这样的，里面没有groups的 151 | def __init__(self,d_in,kernel_size=3): 152 | super().__init__() 153 | self.conv = nn.Sequential( 154 | nn.Conv1d(d_in // 2,d_in // 2,kernel_size=kernel_size,padding=kernel_size // 2,stride=1), 155 | nn.BatchNorm1d(d_in // 2), 156 | nn.Conv1d(d_in // 2,d_in // 2,kernel_size=1,stride=1,padding=0), 157 | nn.SELU(True) 158 | ) 159 | 160 | def forward(self,x): 161 | x1,x2 = x.chunk(2,dim=1) 162 | y = torch.cat((x1,self.conv(x2)),dim=1) 163 | return channel_shuffle(y,2) 164 | 165 | class CBA(nn.Module): 166 | def __init__(self,d_in,d_out,kernel_size,stride=1,groups=1,bias=True,skip=False,act_layer=nn.ReLU): 167 | super().__init__() 168 | padding = kernel_size // 2 169 | self.conv = nn.Conv1d(d_in,d_out,kernel_size=kernel_size,stride=stride,padding=padding,groups=groups,bias=bias) 170 | self.bn = nn.BatchNorm1d(d_out) 171 | # self.bn = GhostBatchNorm1d(d_out) 172 | self.relu = act_layer(True) 173 | self.skip = skip and (stride == 1) and (d_in == d_out) 174 | 175 | def forward(self,x): 176 | identity = x 177 | y = self.relu(self.bn(self.conv(x))) 178 | if self.skip: 179 | y = y + identity 180 | return y 181 | 182 | class DepthwiseSeparableConv(nn.Module): 183 | def __init__(self,d_in,d_out,dw_kernel_size=3,stride=1,skip=True,se_rate=0.2,drop_path_rate=0.,group_size=1,): 184 | super().__init__() 185 | groups = d_in // group_size 186 | padding = dw_kernel_size // 2 187 | self.has_skip = (stride == 1 and d_in == d_out) and skip 188 | self.dw_conv = nn.Conv1d(d_in,d_in,dw_kernel_size,stride=stride,padding=padding,groups=groups) 189 | self.bn1 = nn.BatchNorm1d(d_in) 190 | self.relu = nn.ReLU(inplace=True) 191 | self.se = SEBlock2(d_in,int(d_in * se_rate),act_layer=nn.SELU) if se_rate else nn.Identity() 192 | self.pw_conv = nn.Conv1d(d_in,d_out,1,padding=0) 193 | self.bn2 = nn.BatchNorm1d(d_out) 194 | self.drop_path = nn.Dropout(drop_path_rate) if drop_path_rate > 0. else nn.Identity() 195 | 196 | def forward(self,x): 197 | identity = x 198 | x = self.relu(self.bn1(self.dw_conv(x))) 199 | x = self.se(x) 200 | x = self.relu(self.bn2(self.pw_conv(x))) 201 | if self.has_skip: 202 | x = self.drop_path(x) + identity 203 | return x 204 | 205 | class CAIR(nn.Module): 206 | def __init__(self,d_in,d_out,expand,kernel_size,stride,skip,se_rate,drop_path_rate): 207 | super().__init__() 208 | d_mid = d_in // 2 * expand 209 | self.expand_conv = CBA(d_in // 2,d_mid,kernel_size=1,bias=False) if expand != 1 else nn.Identity() 210 | self.dw_conv = CBA(d_mid,d_mid,kernel_size=kernel_size,stride=stride,groups=d_mid,bias=False) 211 | self.project_conv = nn.Sequential( 212 | nn.Conv1d(d_mid,d_out // 2,kernel_size=1,stride=1,bias=False), 213 | nn.SELU(True) 214 | ) 215 | self.identity_conv = CBA(d_in // 2,d_out // 2,3,stride=2,groups=1,bias=False) if stride == 2 else \ 216 | (nn.Conv1d(d_in // 2,d_out // 2,1) if d_in != d_out else nn.Identity()) 217 | self.se = SEBlock2(d_mid,int(d_mid * se_rate),act_layer=nn.SELU) if se_rate > 0. else nn.Identity() 218 | self.post_conv = CBA(d_out,d_out,3,1,1,act_layer=nn.SELU) 219 | self.skip = (stride == 1 and d_in == d_out) and skip 220 | self.drop_path = nn.Dropout(drop_path_rate) if drop_path_rate > 0. else nn.Identity() 221 | 222 | def forward(self,x): 223 | identity = x.clone() 224 | x1,x2 = x.chunk(2,dim=1) 225 | expand = self.expand_conv(x1) 226 | y1 = self.dw_conv(expand) 227 | y1 = self.se(y1) 228 | y1 = self.project_conv(y1) 229 | y2 = self.identity_conv(x2) 230 | y = torch.cat((y1,y2),dim=1) 231 | y = channel_shuffle(y,2) 232 | y = self.post_conv(y) 233 | y = self.drop_path(y) 234 | return y 235 | 236 | class HFGA(nn.Module): 237 | def __init__(self,d_feat): 238 | super().__init__() 239 | self.dwt = DWT1D(J=1,wave='bior1.1',mode='symmetric') 240 | self.attn_gate = nn.Parameter(torch.Tensor([0.0])) 241 | self.to_q = nn.Conv1d(d_feat,d_feat,1) 242 | self.to_k = nn.Conv1d(d_feat,d_feat,1) 243 | self.to_v = nn.Conv1d(d_feat,d_feat,1) 244 | # self.to_out = nn.Conv1d(d_feat,d_feat,1) 245 | 246 | def compute_attn_matmul(self,q,k,v): # k和v是一样的 247 | # q:(n,c,l1), k & v: (n,c,l2), l1比l2长 248 | attn = k.transpose(1,2) @ q / np.sqrt(q.shape[1]) # (n,l2,l1) 249 | attn = attn - attn.amax(dim=1,keepdim=True).detach() 250 | attn = F.softmax(attn,dim=1) 251 | y = v @ attn # (n,c,l1) 252 | return y 253 | 254 | def forward(self,x): 255 | xl,xh = self.dwt(x) 256 | xh = xh[0] 257 | q = self.to_q(x) 258 | k = self.to_k(xh) 259 | v = self.to_v(xh) 260 | yh = self.compute_attn_matmul(q,k,v) 261 | y = yh * self.attn_gate.tanh() + x 262 | return y 263 | 264 | class LayerNorm(nn.Module): 265 | r""" LayerNorm that supports two data formats: channels_last (default) or channels_first. 266 | The ordering of the dimensions in the inputs. channels_last corresponds to inputs with 267 | shape (batch_size, height, width, channels) while channels_first corresponds to inputs 268 | with shape (batch_size, channels, height, width). 269 | """ 270 | def __init__(self, normalized_shape, eps=1e-6, data_format="channels_last"): 271 | super().__init__() 272 | self.weight = nn.Parameter(torch.ones(normalized_shape)) 273 | self.bias = nn.Parameter(torch.zeros(normalized_shape)) 274 | self.eps = eps 275 | self.data_format = data_format 276 | if self.data_format not in ["channels_last", "channels_first"]: 277 | raise NotImplementedError 278 | self.normalized_shape = (normalized_shape,) 279 | 280 | def forward(self, x): 281 | if self.data_format == "channels_last": 282 | return F.layer_norm(x, self.normalized_shape, self.weight, self.bias, self.eps) 283 | elif self.data_format == "channels_first": 284 | u = x.mean(1, keepdim=True) 285 | s = (x - u).pow(2).mean(1, keepdim=True) 286 | x = (x - u) / torch.sqrt(s + self.eps) 287 | x = self.weight[:, None] * x + self.bias[:, None] 288 | return x 289 | 290 | class ContextBlock(nn.Module): 291 | def __init__(self,d_in,d_hidden,pooling='attn',fusions=['channel_add']): 292 | super().__init__() 293 | self.pooling = pooling 294 | self.conv_mask = nn.Conv1d(d_in,1,kernel_size=1) if pooling == 'attn' else nn.AdaptiveAvgPool1d(1) 295 | if 'channel_add' in fusions: 296 | self.channel_add_conv = nn.Sequential( 297 | nn.Conv1d(d_in,d_hidden,1), 298 | nn.LayerNorm([d_hidden,1]), 299 | nn.ReLU(True), 300 | nn.Conv1d(d_hidden,d_in,1) 301 | ) 302 | else: 303 | self.channel_add_conv = None 304 | if 'channel_mul' in fusions: 305 | self.channel_mul_conv = nn.Sequential( 306 | nn.Conv1d(d_in,d_hidden,1), 307 | nn.LayerNorm([d_hidden,1]), 308 | nn.ReLU(True), 309 | nn.Conv1d(d_hidden,d_in,1) 310 | ) 311 | else: 312 | self.channel_mul_conv = None 313 | self.weights_init() 314 | 315 | def weights_init(self): 316 | if self.pooling == 'attn': 317 | nn.init.kaiming_normal_(self.conv_mask.weight,a=0,mode='fan_in',nonlinearity='relu') 318 | if hasattr(self.conv_mask, 'bias') and self.conv_mask.bias is not None: 319 | nn.init.zeros_(self.conv_mask.bias) 320 | self.conv_mask.inited = True 321 | if self.channel_add_conv is not None: 322 | self.last_zero_init(self.channel_add_conv) 323 | if self.channel_mul_conv is not None: 324 | self.last_zero_init(self.channel_mul_conv) 325 | 326 | def last_zero_init(self,m): 327 | if isinstance(m,nn.Sequential): 328 | nn.init.zeros_(m[-1].weight) 329 | if hasattr(m[-1],'bias') and m[-1].bias is not None: 330 | nn.init.zeros_(m[-1].bias) 331 | else: 332 | nn.init.zeros_(m.weight) 333 | if hasattr(m,'bias') and m.bias is not None: 334 | nn.init.zeros_(m.bias) 335 | 336 | def spatial_pool(self,x): 337 | if self.pooling == 'attn': 338 | context_mask = self.conv_mask(x) # (n,1,l) 339 | context_mask = F.softmax(context_mask,dim=2) # 对l维softmax 340 | context_mask = context_mask.squeeze().unsqueeze(-1) 341 | context = torch.matmul(x,context_mask) # (n,c,l) * (n,l,1) = (n,c,1) 342 | else: 343 | context = self.conv_mask(x) 344 | return context 345 | 346 | def forward(self,x): 347 | context = self.spatial_pool(x) # (n,c,1) 348 | if self.channel_add_conv is not None: 349 | channel_add = self.channel_add_conv(context) 350 | x = x + channel_add 351 | if self.channel_mul_conv is not None: 352 | weights = torch.sigmoid(self.channel_mul_conv(context)) 353 | x = x * weights 354 | return x 355 | 356 | def main(): 357 | ... 358 | 359 | if __name__ == '__main__': 360 | main() -------------------------------------------------------------------------------- /preprocess.py: -------------------------------------------------------------------------------- 1 | 2 | # -*- coding: utf-8 -*- 3 | 4 | import numpy as np 5 | import os,pickle,argparse 6 | from utils import centernorm_size,interpolate_torch 7 | 8 | def preprocess_OLHWDB2(src_root='./data-raw/OLHWDB2',interp=4): 9 | writing = {} 10 | writers = os.listdir(src_root) 11 | for i,w in enumerate(writers): 12 | writing[i] = [] 13 | for file in os.listdir(f'{src_root}/{w}'): 14 | info = [] 15 | with open(f'{src_root}/{w}/{file}','r',encoding='utf-8') as f: 16 | lines = f.readlines() 17 | lines = lines[1:] 18 | lines = [l.strip() for l in lines] 19 | info = [list(map(lambda x:float(x),l.split()[:3])) for l in lines] 20 | info = np.array(info,np.float32) 21 | info = centernorm_size(info) 22 | if interp != None: 23 | info = interpolate_torch(info,interp_ratio=interp) 24 | writing[i].append(info) 25 | tgt_root = src_root.replace('data-raw','data') 26 | os.makedirs(tgt_root,exist_ok=True) 27 | with open(f'{tgt_root}/OLHWDB2.pkl','wb') as f: 28 | pickle.dump(writing,f) 29 | 30 | def preprocess_DCOHE(src_root='./data-raw/DCOH-E'): 31 | writing = {} 32 | writers = os.listdir(src_root) 33 | for i,w in enumerate(writers): 34 | writing[w] = [] 35 | for file in os.listdir(f'{src_root}/{w}'): 36 | info = [] 37 | with open(f'{src_root}/{w}/{file}','r',encoding='utf-8') as f: 38 | lines = f.readlines() 39 | lines = lines[1:] 40 | lines = [l.strip() for l in lines] 41 | info = [list(map(lambda x:float(x),l.split()[:3])) for l in lines] 42 | info = np.array(info,np.float32) 43 | info = centernorm_size(info) 44 | if 'dcoh-e' in file: 45 | info = interpolate_torch(info,interp_ratio=2) 46 | writing[w].append(info) 47 | tgt_root = src_root.replace('data-raw','data') 48 | os.makedirs(tgt_root,exist_ok=True) 49 | with open(f'{tgt_root}/DCOH-E.pkl','wb') as f: 50 | pickle.dump(writing,f) 51 | 52 | def preprocess_COUCH(src_root='./data-raw/COUCH09',interp=4): 53 | writing = {} 54 | writers = os.listdir(src_root) 55 | for i,w in enumerate(writers): 56 | writing[i] = [] 57 | for file in os.listdir(f'{src_root}/{w}'): 58 | with open(f'{src_root}/{w}/{file}','r',encoding='utf-8') as f: 59 | lines = f.readlines() 60 | lines = lines[1:] 61 | lines = [l.strip() for l in lines] 62 | info = [list(map(lambda x:float(x),l.split()[:3])) for l in lines] 63 | info = np.array(info,np.float32) 64 | info = centernorm_size(info) 65 | if interp != None: 66 | info = interpolate_torch(info,interp_ratio=interp) 67 | writing[i].append(info) 68 | tgt_root = src_root.replace('data-raw','data') 69 | os.makedirs(tgt_root,exist_ok=True) 70 | with open(f'{tgt_root}/COUCH09.pkl','wb') as f: 71 | pickle.dump(writing,f) 72 | 73 | if __name__ == '__main__': 74 | parser = argparse.ArgumentParser() 75 | parser.add_argument('--dataset',type=str,default='olhwdb2',help='processed dataset names: [olhwdb2,dcohe,couch]') 76 | opt = parser.parse_args() 77 | func = globals()[f'preprocess_{opt.dataset.upper()}'] 78 | print(f'start preprocessing {opt.dataset.upper()} dataset.') 79 | func() 80 | print(f'end preprocessing {opt.dataset.upper()} dataset.') 81 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | absl-py==2.1.0 2 | aiohttp==3.9.5 3 | aiosignal==1.3.1 4 | albumentations==1.4.3 5 | antlr4-python3-runtime==4.9.3 6 | asttokens @ file:///home/conda/feedstock_root/build_artifacts/asttokens_1698341106958/work 7 | async-timeout==4.0.3 8 | attrs==23.2.0 9 | backcall @ file:///home/conda/feedstock_root/build_artifacts/backcall_1592338393461/work 10 | beartype==0.17.2 11 | beautifulsoup4==4.12.3 12 | cachetools==5.3.1 13 | causal-conv1d==1.3.0.post1 14 | certifi==2023.7.22 15 | charset-normalizer==3.2.0 16 | colorama==0.4.6 17 | comm @ file:///home/conda/feedstock_root/build_artifacts/comm_1710320294760/work 18 | contourpy==1.1.0 19 | coolgpus==0.23 20 | cycler==0.11.0 21 | debugpy @ file:///croot/debugpy_1690905042057/work 22 | decorator @ file:///home/conda/feedstock_root/build_artifacts/decorator_1641555617451/work 23 | dominate==2.9.1 24 | easydict==1.11 25 | edit-distance==1.0.6 26 | einops==0.7.0 27 | ema-pytorch==0.4.3 28 | entrypoints @ file:///home/conda/feedstock_root/build_artifacts/entrypoints_1643888246732/work 29 | executing @ file:///home/conda/feedstock_root/build_artifacts/executing_1698579936712/work 30 | fastdtw==0.3.4 31 | filelock==3.12.2 32 | fonttools==4.42.0 33 | frozenlist==1.4.1 34 | fsspec==2023.6.0 35 | google-auth==2.29.0 36 | google-auth-oauthlib==1.0.0 37 | grpcio==1.62.1 38 | huggingface-hub==0.23.4 39 | idna==3.4 40 | iisignature @ file:///home/WD16T/zpr/iisignature-0.24-cp38-cp38-linux_x86_64.whl#sha256=24b2e81340e26b1a8501d721b87892c706fdcb98c4ee72d24721b76602ae80ec 41 | imagecorruptions==1.1.2 42 | imageio==2.34.0 43 | imgaug==0.4.0 44 | importlib-metadata==6.8.0 45 | importlib-resources==6.0.0 46 | ipykernel @ file:///home/conda/feedstock_root/build_artifacts/ipykernel_1708996548741/work 47 | ipython @ file:///home/conda/feedstock_root/build_artifacts/ipython_1680185408135/work 48 | jedi @ file:///home/conda/feedstock_root/build_artifacts/jedi_1696326070614/work 49 | joblib==1.3.1 50 | jsonpatch==1.33 51 | jsonpointer==2.4 52 | jupyter-client @ file:///home/conda/feedstock_root/build_artifacts/jupyter_client_1654730843242/work 53 | jupyter_core @ file:///home/conda/feedstock_root/build_artifacts/jupyter_core_1710257397447/work 54 | kiwisolver==1.4.4 55 | kornia==0.7.2 56 | kornia_rs==0.1.3 57 | lazy_loader==0.4 58 | lightning-utilities==0.11.2 59 | llvmlite==0.40.1 60 | lmdb==1.4.1 61 | mamba-ssm==2.1.0 62 | Markdown==3.6 63 | MarkupSafe==2.1.5 64 | matplotlib==3.7.2 65 | matplotlib-inline @ file:///home/conda/feedstock_root/build_artifacts/matplotlib-inline_1660814786464/work 66 | multidict==6.0.5 67 | natsort==8.4.0 68 | nest_asyncio @ file:///home/conda/feedstock_root/build_artifacts/nest-asyncio_1705850609492/work 69 | networkx==3.1 70 | ninja==1.11.1.1 71 | numba==0.57.1 72 | numpy==1.24.4 73 | nvidia-ml-py==12.535.77 74 | nvitop==1.2.0 75 | oauthlib==3.2.2 76 | omegaconf==2.3.0 77 | opencv-contrib-python==4.9.0.80 78 | opencv-python==4.9.0.80 79 | opencv-python-headless==4.9.0.80 80 | packaging==23.1 81 | pandas==2.0.3 82 | parso @ file:///home/conda/feedstock_root/build_artifacts/parso_1638334955874/work 83 | pexpect @ file:///home/conda/feedstock_root/build_artifacts/pexpect_1706113125309/work 84 | pickleshare @ file:///home/conda/feedstock_root/build_artifacts/pickleshare_1602536217715/work 85 | Pillow==10.0.0 86 | platformdirs @ file:///home/conda/feedstock_root/build_artifacts/platformdirs_1706713388748/work 87 | pretty-errors==1.2.25 88 | prompt-toolkit @ file:///home/conda/feedstock_root/build_artifacts/prompt-toolkit_1702399386289/work 89 | protobuf==5.26.0 90 | psutil @ file:///opt/conda/conda-bld/psutil_1656431268089/work 91 | ptflops==0.7 92 | ptyprocess @ file:///home/conda/feedstock_root/build_artifacts/ptyprocess_1609419310487/work/dist/ptyprocess-0.7.0-py2.py3-none-any.whl 93 | pure-eval @ file:///home/conda/feedstock_root/build_artifacts/pure_eval_1642875951954/work 94 | pyasn1==0.5.1 95 | pyasn1-modules==0.3.0 96 | Pygments @ file:///home/conda/feedstock_root/build_artifacts/pygments_1700607939962/work 97 | pyparsing==3.0.9 98 | python-dateutil==2.8.2 99 | pytorch-lightning==2.0.0 100 | pytorch-metric-learning==2.3.0 101 | pytorch-wavelets==1.3.0 102 | pytz==2023.3 103 | PyWavelets==1.4.1 104 | PyYAML==6.0.1 105 | pyzmq @ file:///croot/pyzmq_1705605076900/work 106 | regex==2024.5.15 107 | requests==2.31.0 108 | requests-oauthlib==2.0.0 109 | rsa==4.9 110 | safetensors==0.4.3 111 | scikit-image==0.21.0 112 | scikit-learn==1.3.2 113 | scipy==1.10.1 114 | seaborn==0.13.2 115 | shapely==2.0.3 116 | six @ file:///home/conda/feedstock_root/build_artifacts/six_1620240208055/work 117 | soupsieve==2.6 118 | stack-data @ file:///home/conda/feedstock_root/build_artifacts/stack_data_1669632077133/work 119 | tensorboard==2.14.0 120 | tensorboard-data-server==0.7.2 121 | termcolor==2.3.0 122 | thop==0.1.1.post2209072238 123 | threadpoolctl==3.2.0 124 | tifffile==2023.7.10 125 | timm==0.3.2 126 | tokenizers==0.19.1 127 | torch==1.12.1+cu113 128 | torchaudio==0.12.1+cu113 129 | torchmetrics==1.4.0 130 | torchnet==0.0.4 131 | torchstat==0.0.7 132 | torchvision==0.13.1+cu113 133 | tornado @ file:///home/conda/feedstock_root/build_artifacts/tornado_1648827257044/work 134 | tqdm==4.65.0 135 | traitlets @ file:///home/conda/feedstock_root/build_artifacts/traitlets_1710254411456/work 136 | transformers==4.41.2 137 | triton==2.3.1 138 | tslearn==0.6.1 139 | typing_extensions @ file:///home/conda/feedstock_root/build_artifacts/typing_extensions_1708904622550/work 140 | tzdata==2023.3 141 | urllib3==2.0.4 142 | visdom==0.2.4 143 | wcwidth @ file:///home/conda/feedstock_root/build_artifacts/wcwidth_1704731205417/work 144 | websocket-client==1.7.0 145 | Werkzeug==3.0.1 146 | yarl==1.9.4 147 | zipp==3.16.2 148 | -------------------------------------------------------------------------------- /test.py: -------------------------------------------------------------------------------- 1 | # -*- coding:utf-8 -*- 2 | 3 | import torch 4 | from torch.utils.data import DataLoader 5 | from torch.backends import cudnn 6 | import argparse,os,time,json 7 | from model import DOLPHIN 8 | from dataset import Writing,collate_fn 9 | from utils import create_logger,load_ckpt,l2_norm,fuse_all_conv_bn 10 | import numpy as np 11 | from evaluate import compute_metrics 12 | import pickle 13 | from natsort import natsorted 14 | import matplotlib.pyplot as plt 15 | from ptflops import get_model_complexity_info 16 | from thop import profile 17 | from torchstat import stat 18 | 19 | torch._C._jit_set_profiling_mode(False) 20 | torch._C._jit_set_profiling_executor(False) 21 | torch.cuda.empty_cache() 22 | 23 | parser = argparse.ArgumentParser() 24 | parser.add_argument('--batch_size',type=int,default=8) 25 | parser.add_argument('--num_classes',type=int,default=1731) 26 | parser.add_argument('--epoch',type=int,default=80) 27 | parser.add_argument('--seed',type=int,default=123) 28 | parser.add_argument('--cuda',type=bool,default=True) 29 | parser.add_argument('--folder',type=str,default='./data/OLIWER') 30 | parser.add_argument('--ngpu',type=int,default=1) 31 | parser.add_argument('--gpu',type=str,default='0') 32 | parser.add_argument('--weights',type=str,default='./weights') 33 | parser.add_argument('--output_root',type=str,default='./output') 34 | parser.add_argument('--log_root',type=str,default='./logs') 35 | parser.add_argument('--dba',action='store_true') 36 | parser.add_argument('--rerank',action='store_true') 37 | parser.add_argument('--name',type=str,default='DOLPHIN') 38 | opt = parser.parse_args() 39 | 40 | # with open(f'{opt.weights}/settings.json','r',encoding='utf-8') as f: 41 | # settings = json.loads(f.read()) 42 | # opt.seed = settings['seed'] 43 | # opt.name = settings['name'] 44 | # opt.notes = settings['notes'] 45 | # opt.log_root = settings['log_root'] 46 | # # opt.folder = settings['folder'] 47 | # # opt.gpu = settings['gpu'] 48 | 49 | np.random.seed(opt.seed) 50 | torch.manual_seed(opt.seed) 51 | torch.cuda.manual_seed_all(opt.seed) 52 | 53 | logger = create_logger(opt.log_root,name=opt.name,test=True) 54 | 55 | # query_root = f'{opt.folder}/query-tf.pkl' 56 | # with open(query_root,'rb') as f: 57 | # query_data = pickle.load(f,encoding='iso-8859-1') 58 | gallery_root = f'{opt.folder}/test-tf.pkl' 59 | with open(gallery_root,'rb') as f: 60 | gallery_data = pickle.load(f,encoding='iso-8859-1') 61 | # handwriting_info = {} 62 | # min_sample = 10000 63 | # for k in query_data: 64 | # handwriting_info[k] = query_data[k] + gallery_data[k] 65 | # min_sample = min(len(handwriting_info[k]),min_sample) 66 | # print(min_sample,len(query_data),len(gallery_data)) 67 | gallery_dataset = Writing(gallery_data,train=False) 68 | d_in = gallery_dataset.feature_dims 69 | 70 | gallery_loader = DataLoader(gallery_dataset,batch_size=opt.batch_size,shuffle=False,collate_fn=collate_fn) 71 | 72 | model = DOLPHIN(d_in,opt.num_classes) 73 | 74 | if opt.cuda and torch.cuda.is_available(): 75 | torch.cuda.set_device(int(opt.gpu)) 76 | device = torch.device(f'cuda:{opt.gpu}') 77 | else: 78 | device = torch.device('cpu') 79 | model = model.to(device) 80 | 81 | logger.info(f'\ngallery root: {gallery_root}\n' 82 | f'gallery loader length: {len(gallery_loader)} gallery features length: {len(gallery_dataset)}\n' 83 | f'model: {model.__class__.__name__}\nDBA & AQE: {opt.dba}\nRerank: {opt.rerank}') 84 | 85 | def extract_features(model,data_loader,time_model): 86 | for i,(x,features_lens,user_labels) in enumerate(data_loader): 87 | x = torch.from_numpy(x).to(device) 88 | features_lens = torch.tensor(features_lens).long().to(device) 89 | user_labels = torch.from_numpy(user_labels).long() 90 | s = time.time() 91 | y_vector = model(x,features_lens)[0] 92 | # y_vector = model(x)[0] 93 | e = time.time() 94 | time_model += (e - s) 95 | y_vector = l2_norm(y_vector) 96 | if i == 0: 97 | features = torch.zeros(len(data_loader.dataset),y_vector.shape[1]) 98 | start = i * opt.batch_size 99 | end = min((i + 1) * opt.batch_size,len(data_loader.dataset)) 100 | features[start:end,:] = y_vector 101 | if i == 0: 102 | labels = user_labels 103 | else: 104 | labels = torch.cat([labels,user_labels],0) 105 | return features.cpu().numpy(),labels.cpu().numpy(),time_model 106 | 107 | def transform_user2feat(features,labels): 108 | label_indices = natsorted(np.unique(labels)) 109 | user2feat = {} 110 | for i in label_indices: 111 | pos = np.where(labels == i)[0] 112 | user2feat[i] = features[pos] 113 | return user2feat 114 | 115 | @torch.no_grad() 116 | def test_impl(model): 117 | model = model.eval() 118 | model = model.to(device) 119 | # model = fuse_all_conv_bn(model) 120 | 121 | time_elapsed_start = time.time() 122 | all_features,all_labels,time_model = extract_features(model,gallery_loader,0) 123 | user2feat = transform_user2feat(all_features,all_labels) 124 | repeat_times = 1 125 | logger.info(f'repeat times: {repeat_times}') 126 | gallery_labels,query_labels = [],[] 127 | for i in natsorted(np.unique(all_labels)): 128 | gallery_labels.extend([i] * (len(user2feat[i]) - 1)) 129 | query_labels.append(i) 130 | gallery_labels = np.array(gallery_labels) 131 | query_labels = np.array(query_labels) 132 | aps,top1s,top5s,top10s = [],[],[],[] 133 | for _ in range(repeat_times): 134 | gallery_features,query_features = [],[] 135 | label_indices = natsorted(np.unique(all_labels)) 136 | for i in label_indices: 137 | idx = np.random.choice(len(user2feat[i]),size=1)[0] 138 | gallery_features.append(user2feat[i][:idx]) 139 | gallery_features.append(user2feat[i][idx + 1:]) 140 | query_features.append(user2feat[i][idx]) 141 | gallery_features = np.concatenate(gallery_features) 142 | query_features = np.array(query_features) 143 | res = { 144 | 'gallery_feature':gallery_features,'gallery_label':gallery_labels, 145 | 'query_feature':query_features,'query_label':query_labels, 146 | } 147 | _,ap,top1,top5,top10 = compute_metrics(res,logger,opt.dba,device,verbose=False) 148 | aps.append(ap) 149 | top1s.append(top1) 150 | top5s.append(top5) 151 | top10s.append(top10) 152 | ap_mean,ap_std = np.mean(aps),np.std(aps) 153 | top1_mean,top1_std = np.mean(top1s),np.std(top1s) 154 | top5_mean,top5_std = np.mean(top5s),np.std(top5s) 155 | top10_mean,top10_std = np.mean(top10s),np.std(top10s) 156 | logger.info(f'[final] Rank@1: {top1_mean:.4f}% ({top1_std:.4f}%) Rank@5: {top5_mean:.4f}% ({top5_std:.4f}%) ' 157 | f'Rank@10: {top10_mean:.4f}% ({top10_std:.4f}%)') 158 | logger.info(f'[final] mAP: {ap_mean * 100.:.4f}% ({ap_std * 100:.4f}%)') 159 | logger.info(f'time elapsed: {time.time() - time_elapsed_start:.5f}s\n') 160 | 161 | def test(): 162 | load_ckpt(model,opt.weights,device,logger,mode='test') 163 | test_impl(model) 164 | 165 | def main(): 166 | test() 167 | 168 | if __name__ == '__main__': 169 | main() 170 | -------------------------------------------------------------------------------- /utils.py: -------------------------------------------------------------------------------- 1 | # -*- coding:utf-8 -*- 2 | 3 | from scipy import signal 4 | from torch.nn.utils import fuse_conv_bn_eval 5 | import numpy as np 6 | import cv2,logging,sys,time,os 7 | from functools import wraps,lru_cache,reduce 8 | from termcolor import colored 9 | import torch 10 | import torch.nn as nn 11 | import torch.nn.functional as F 12 | import matplotlib.pyplot as plt 13 | from PIL import Image 14 | 15 | class ButterWorthLPF: 16 | # 巴特沃斯低通滤波器 17 | def __init__(self,order=3,half_pnt=15.0,fnyquist=100.0): 18 | ''' 19 | scipy.signal.butter(N, Wn, btype='low', analog=False, output='ba', fs=None) 20 | 这个是滤波器，高通低通带通 21 | N是滤波器阶数，Wn是3dB带宽点，dB就是radians/frequency(弧度/频率) 22 | 3dB就是功率下降到二分之一的点，所以这里叫做half_pnt(point) 23 | byte决定是什么通，analog=False表示模拟滤波器，True表示数字滤波器 24 | 'ba'表示输出分子和分母的系数;'zpk'表示输出零极点;'sos'表示输出second-order sections. 25 | 默认是'ba'，意思应该是系统响应函数分子上的系数bk和分母上的系数ak 26 | https://wenku.baidu.com/view/adec241352d380eb62946d82.html 27 | order: 滤波器阶数 28 | half_pnt: 3dB点，功率降到一半的点 29 | fnyquist: sampling frequency，奈奎斯特采样率，采样频率必须要大于原始频率的两倍，书上叫fs 30 | ''' 31 | fM = 0.5 * fnyquist # fM是原始频率，命名参考信号与系统第2版第七章 32 | half_pnt /= fM 33 | b,a = signal.butter(order,half_pnt,'low') 34 | self.b = b # 分子 35 | self.a = a # 分母 36 | 37 | def __call__(self,x): # x就是输入进来要滤波的数据 38 | return signal.filtfilt(self.b,self.a,x) 39 | # 将data通过零相位滤波器，零相位的意思就是输入和输出信号的相位完全相同，相移为0 40 | # 至于为什么是零相位滤波，暂时只能说经验值 41 | 42 | lpf = ButterWorthLPF() 43 | 44 | def difference(x): # 差分，两跨点之间相减 45 | ''' 46 | numpy.convolve(a, v, mode='full')，a的长度是N，v是M 47 | mode可以取'full','same','valid',full的意思是长度为N+M-1，same的意思是长度为max(M,N)， 48 | valid的意思是长度为max(M,N) - min(M,N) + 1，只有完全重叠的点有效， 49 | 边缘点无效（边缘点就是有一个序列突出去的那些） 50 | mode='same'的时候，比如[5,8,6,9,1,2]*[0.5,0,-0.5]，full的长度是8，然后两边减一个， 51 | 刚刚好就是6了，而且[0.5,0,-0.5]是会反转的，[4., 0.5, 0.5, -2.5, -3.5, -0.5] 52 | delta_x[0] = delta_x[1] 53 | delta_x[-1] = delta_x[-2] 54 | 这两句的作用就是将4代替为0.5，-0.5代替为-3.5，是因为那两个点没人减 55 | ''' 56 | delta_x = np.convolve(x,[0.5,0,-0.5],mode='same') 57 | delta_x[0] = delta_x[1] 58 | delta_x[-1] = delta_x[-2] 59 | return delta_x 60 | 61 | def difference_theta(x): # 输入的x是角度 62 | delta_x = np.zeros_like(x) 63 | delta_x[1:-1] = x[2:] - x[:-2] 64 | delta_x[-1] = delta_x[-2] 65 | delta_x[0] = delta_x[1] 66 | t = np.where(np.abs(delta_x) > np.pi) 67 | delta_x[t] = np.sign(delta_x[t]) * 2 * np.pi 68 | delta_x *= 0.5 69 | return delta_x 70 | 71 | def extract_features(handwritings,features,gpnoise=None,num=2,transform=False): 72 | ''' 73 | paths: 路径列表，第一维应该是点的个数 74 | features: 这个就是所有特征的列表，是单个feature append进去的 75 | num: 使用的信息个数，比如x,y,pressure...,2就是只用012前三个 76 | gpnoise: 不知道 77 | transform: 不知道 78 | use_finger: 是否是用手指写的 79 | ''' 80 | for handwriting in handwritings: 81 | pressure = handwriting[:,num] 82 | handwriting = handwriting[:,0:num] # (x,y,pressure) 83 | handwriting[:,0] = lpf(handwriting[:,0]) 84 | handwriting[:,1] = lpf(handwriting[:,1]) 85 | delta_x = difference(handwriting[:,0]) 86 | delta_y = difference(handwriting[:,1]) 87 | v = np.sqrt(delta_x ** 2 + delta_y ** 2) # 速度 88 | theta = np.arctan2(delta_y,delta_x) 89 | cos_theta = np.cos(theta) 90 | sin_theta = np.sin(theta) 91 | delta_v = difference(v) 92 | delta_theta = np.abs(difference_theta(theta)) 93 | log_curve_radius = np.log((v + 0.05) / (delta_theta + 0.05)) # log的曲线弧度 94 | delta_v2 = np.abs(v * delta_theta) 95 | acceleration = np.sqrt(delta_v ** 2 + delta_v2 ** 2) 96 | 97 | # None在这里的作用是升维，比如说[2,2]会变成[2,1,2],concat起来就是[2,x,2] 98 | single_feature = np.concatenate((delta_x[:,None],delta_y[:,None],v[:,None], 99 | cos_theta[:,None],sin_theta[:,None],theta[:,None],log_curve_radius[:,None], 100 | acceleration[:,None],delta_v[:,None],delta_v2[:,None],delta_theta[:,None], 101 | pressure[:,None]),axis=1).astype(np.float32) 102 | single_feature[:,:-1] = (single_feature[:,:-1] - np.mean(single_feature[:,:-1],axis=0)) / \ 103 | np.std(single_feature[:,:-1],axis=0) 104 | # single_feature[:,:-1] = regression_based_norm(single_feature[:,:-1]) 105 | features.append(single_feature) 106 | 107 | def time_functions(handwriting,num=2): 108 | # handwriting = deepcopy(handwriting_org) 109 | pressure = handwriting[:,num] 110 | # pressure = np.ones_like(pressure) 111 | handwriting = handwriting[:,0:num] # (x,y,pressure) 112 | handwriting[:,0] = lpf(handwriting[:,0]) 113 | handwriting[:,1] = lpf(handwriting[:,1]) 114 | delta_x = difference(handwriting[:,0]) 115 | delta_y = difference(handwriting[:,1]) 116 | v = np.sqrt(delta_x ** 2 + delta_y ** 2) # 速度 117 | theta = np.arctan2(delta_y,delta_x) 118 | cos_theta = np.cos(theta) 119 | sin_theta = np.sin(theta) 120 | delta_v = difference(v) 121 | delta_theta = np.abs(difference_theta(theta)) 122 | log_curve_radius = np.log((v + 0.05) / (delta_theta + 0.05)) # log的曲线弧度 123 | delta_v2 = np.abs(v * delta_theta) 124 | acceleration = np.sqrt(delta_v ** 2 + delta_v2 ** 2) 125 | delta_x2 = difference(delta_x) 126 | delta_y2 = difference(delta_y) 127 | # None在这里的作用是升维，比如说[2,2]会变成[2,1,2],concat起来就是[2,x,2] 128 | single_feature = np.concatenate((delta_x[:,None],delta_y[:,None],delta_x2[:,None],delta_y2[:,None],v[:,None], 129 | cos_theta[:,None],sin_theta[:,None],theta[:,None],log_curve_radius[:,None], 130 | acceleration[:,None],delta_v[:,None],delta_v2[:,None],delta_theta[:,None], 131 | pressure[:,None]),axis=1).astype(np.float32) 132 | 133 | single_feature[:,:-1] = (single_feature[:,:-1] - np.mean(single_feature[:,:-1],axis=0)) / np.std(single_feature[:,:-1],axis=0) 134 | return single_feature 135 | 136 | def letterbox_image(img,target_h,target_w): 137 | img_h,img_w = img.shape 138 | scale = min(target_h / img_h,target_w / img_w) 139 | # 长宽比目标size小的，可以变大，不过变大不一定有必要 140 | new_w,new_h = int(img_w * scale),int(img_h * scale) # 这样做就依然保持了长宽比 141 | img = cv2.resize(img,(new_w,new_h),interpolation=cv2.INTER_AREA) 142 | new_img = np.ones((target_h,target_w),dtype=np.uint8) * 255 143 | up = (target_h - new_h) // 2 144 | left = (target_w - new_w) // 2 145 | new_img[up:up + new_h,left:left + new_w] = img 146 | return new_img 147 | 148 | def interpolate_torch(org_info,interp_ratio): 149 | l = len(org_info) 150 | org_info = torch.tensor(org_info).view(1,1,l,-1) 151 | new_info = F.interpolate(org_info,size=(l * interp_ratio,3),mode='bicubic').squeeze().numpy() 152 | return new_info 153 | 154 | def load_ckpt(model,pretrained_root,device,logger,optimizer=None,scheduler=None,mode='train',resume=False): 155 | # pretrained=True是否基于其他任务的预训练 156 | state_dict = torch.load(pretrained_root,map_location=device) 157 | if mode == 'train': 158 | if resume: 159 | optimizer.load_state_dict(state_dict['optimizer']) 160 | scheduler.load_state_dict(state_dict['lr_scheduler']) 161 | print(model.load_state_dict(state_dict['model'])) 162 | start_epoch = state_dict['epoch'] + 1 163 | logger.info(f'mode: "{mode} + resume" {pretrained_root} successfully loaded.') 164 | else: 165 | state_dict = state_dict['model'] if 'model' in state_dict else state_dict 166 | state_dict = {k:v for k,v in state_dict.items() if k in model.state_dict().keys() and v.numel() == model.state_dict()[k].numel()} 167 | print(model.load_state_dict(state_dict,strict=False)) 168 | logger.info(f'mode: "{mode} + pretrained" {pretrained_root} successfully loaded.') 169 | start_epoch = 0 170 | return start_epoch 171 | else: 172 | state_dict = state_dict['model'] if 'model' in state_dict else state_dict 173 | state_dict = {k:v for k,v in state_dict.items() if k in model.state_dict().keys() and v.numel() == model.state_dict()[k].numel()} 174 | print(model.load_state_dict(state_dict,strict=False)) 175 | # model.load_state_dict(state_dict['model']) 176 | logger.info(f'mode: "{mode}" {pretrained_root} successfully loaded.') 177 | 178 | @lru_cache() 179 | def create_logger(log_root,name='',test=False): 180 | os.makedirs(f'{log_root}',exist_ok=True) 181 | logger = logging.getLogger(name) 182 | logger.setLevel(logging.INFO) 183 | logger.propagate = False 184 | 185 | color_fmt = colored('[%(asctime)s %(name)s]','green') + \ 186 | colored('(%(filename)s %(lineno)d)','yellow') + ': %(levelname)s %(message)s' 187 | console_handler = logging.StreamHandler(sys.stderr) 188 | console_handler.setLevel(logging.INFO) # 分布式的等级 189 | console_handler.setFormatter(logging.Formatter(fmt=color_fmt,datefmt='%Y-%m-%d %H:%M:%S')) 190 | logger.addHandler(console_handler) 191 | 192 | fmt = '[%(asctime)s %(name)s] (%(filename)s %(lineno)d): %(levelname)s %(message)s' 193 | date = time.strftime('%Y-%m-%d') if not test else time.strftime('%Y-%m-%d') + '-test' 194 | file_handler = logging.FileHandler(f'{log_root}/log-{date}.txt',mode='a') 195 | file_handler.setLevel(logging.INFO) 196 | file_handler.setFormatter(logging.Formatter(fmt=fmt,datefmt='%Y-%m-%d %H:%M:%S')) 197 | logger.addHandler(file_handler) 198 | return logger 199 | 200 | def l2_norm(x): # x:(batch_size,seq_len) 201 | org_size = x.size() 202 | x_pow = torch.pow(x,2) 203 | x_pow = torch.sum(x_pow,1).add_(1e-6) 204 | x_pow = torch.sqrt(x_pow) 205 | y = torch.div(x,x_pow.view(-1,1).expand_as(x)).view(org_size) 206 | return y 207 | 208 | def centernorm_size(handwriting,coord_idx=[0,1]): 209 | # coord_idx其实是下标，就是说在handwriting这个二维数组里面是下标0和1分别是x和y 210 | assert len(coord_idx) == 2 211 | pos = handwriting[:,coord_idx] 212 | minx = np.min(pos,axis=0) 213 | maxn = np.max(pos,axis=0) 214 | pos = (pos - (maxn + minx) / 2.) / np.max(maxn - minx) # 不知道为什么这样除，经验值 215 | handwriting[:,coord_idx] = pos 216 | return handwriting 217 | 218 | def norm_pressure(handwriting,pressure_idx=2): # 单纯变0到1，但是其实可以不用 219 | pressure = handwriting[:,pressure_idx] 220 | maxn = np.max(pressure) 221 | pressure /= maxn 222 | handwriting[:,pressure_idx] = pressure 223 | return handwriting 224 | 225 | def fuse_all_conv_bn(model): 226 | stack = [] 227 | for name,module in model.named_children(): 228 | if list(module.named_children()): 229 | fuse_all_conv_bn(module) 230 | if isinstance(module,nn.BatchNorm1d): 231 | if not stack: # 空的 232 | continue 233 | if isinstance(stack[-1][1],nn.Conv1d): 234 | setattr(model,stack[-1][0],fuse_conv_bn_eval(stack[-1][1],module)) 235 | setattr(model,name,nn.Identity()) 236 | else: 237 | stack.append((name,module)) 238 | return model 239 | 240 | def extract_and_store(src_root='./data/OCNOLHW-granular/train.pkl'): 241 | import pickle 242 | with open(src_root,'rb') as f: 243 | handwriting_info = pickle.load(f,encoding='iso-8859-1') 244 | writing = {} 245 | print(len(handwriting_info)) 246 | for i,k in enumerate(handwriting_info.keys()): 247 | writing[k] = [] 248 | a = time.time() 249 | for each in handwriting_info[k]: 250 | writing[k].append(time_functions(each)) 251 | print(time.time() - a) 252 | break 253 | 254 | def clock(func): 255 | @wraps(func) 256 | def impl(*args,**kwargs): 257 | start = time.perf_counter() 258 | res = func(*args,**kwargs) 259 | end = time.perf_counter() 260 | args_list = [] 261 | if args: 262 | args_list.extend([repr(arg) for arg in args]) 263 | if kwargs: 264 | args_list.extend([f'{key}={value}' for key,value in kwargs.items()]) 265 | args_str = ','.join(i for i in args_list) 266 | print(f'[executed in {(end - start):.5f}s, ' 267 | f'{func.__name__}({args_str}) -> {res}]') 268 | return res 269 | return impl 270 | 271 | def main(): 272 | ... 273 | 274 | if __name__ == '__main__': 275 | main() -------------------------------------------------------------------------------- /weights/model.pth: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/SCUT-DLVCLab/DOLPHIN/fc90f848ef11cbdcc2f89fabf3a9ed68343641d0/weights/model.pth --------------------------------------------------------------------------------