├── .gitignore ├── LICENSE ├── README.md ├── config.py ├── experiments ├── spatial_transfer.ipynb └── training.ipynb ├── models └── dcm.py ├── preprocessing ├── preprocess_ARD.ipynb └── preprocess_CDL.ipynb ├── requirements.txt └── utils ├── date.py ├── helper.py ├── io_func.py ├── logger.py └── timer.py /.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 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow 98 | __pypackages__/ 99 | 100 | # Celery stuff 101 | celerybeat-schedule 102 | celerybeat.pid 103 | 104 | # SageMath parsed files 105 | *.sage.py 106 | 107 | # Environments 108 | .env 109 | .venv 110 | env/ 111 | venv/ 112 | ENV/ 113 | env.bak/ 114 | venv.bak/ 115 | 116 | # Spyder project settings 117 | .spyderproject 118 | .spyproject 119 | 120 | # Rope project settings 121 | .ropeproject 122 | 123 | # mkdocs documentation 124 | /site 125 | 126 | # mypy 127 | .mypy_cache/ 128 | .dmypy.json 129 | dmypy.json 130 | 131 | # Pyre type checker 132 | .pyre/ 133 | 134 | # pytype static type analyzer 135 | .pytype/ 136 | 137 | # Cython debug symbols 138 | cython_debug/ -------------------------------------------------------------------------------- /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 | . -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # DeepCropMapping: A multi-temporal deep learning approach with improved spatial generalizability for dynamic corn and soybean mapping 2 | 3 | This repository is the official implementation of DeepCropMapping: A multi-temporal deep learning approach with improved spatial generalizability for dynamic corn and soybean mapping. 4 | 5 | ## Requirements 6 | 7 | - torch 8 | - numpy 9 | - pandas 10 | - scikit-learn 11 | - jupyter 12 | 13 | The code has been tested in the following environment: 14 | Ubuntu 16.04.4 LTS, Python 3.5.2, PyTorch 1.2.0 15 | 16 | ## Data 17 | 18 | The preprocessed data (`.npy` files) for model training and evaluation is not directly provided here due to the large data volume. You can download raw Landsat Analysis Ready Data (ARD) from [EarthExplore](https://earthexplorer.usgs.gov/) and raw Cropland Data Layer (CDL) from [CropScape](https://nassgeodata.gmu.edu/CropScape/), then follow the code in the `preprocessing` folder to generate the `.npy` files. The raw Landsat ARD and CDL data should be stored in a new `data` folder that has the following structure (specific downloaded file names may change): 19 | 20 | ``` 21 | data 22 | ├── Site_A 23 | │   ├── ARD 24 | │   │   ├── 2015 25 | │   │   │   ├── LC08_CU_018007_20150424_20181206_C01_V01_PIXELQA.tif 26 | │   │   │   ├── LC08_CU_018007_20150424_20181206_C01_V01_SRB2.tif 27 | │   │   │   └── . . . 28 | │   │   ├── . . . 29 | │   │   └── 2018 30 | │   └── CDL 31 | │   ├── CDL_2015_clip_20190409130240_375669680.tif 32 | │   ├── . . . 33 | │   └── CDL_2018_clip_20190409125506_12566268.tif 34 | ├── Site_B 35 | ├── . . . 36 | └── Site_F 37 | ``` 38 | 39 | The preprocessed data should be stored in the `preprocessing/out` folder that has the following structure: 40 | 41 | ``` 42 | preprocessing/out 43 | ├── Site_A 44 | │   ├── x-2015.npy 45 | │   ├── y-2015.npy 46 | │   ├── . . . 47 | │   ├── x-2018.npy 48 | │   └── y-2018.npy 49 | ├── Site_B 50 | ├── . . . 51 | └── Site_F 52 | ``` 53 | 54 | ## Training and evaluation 55 | 56 | - The PyTorch implementation of DeepCropMapping (DCM) model is located in the `models` folder. 57 | - The `utils` folder contains some utilities that are used for data loading, normalization, training and evluation. 58 | 59 | The specific training and evaluation process can be executed by running the `.ipynb` files in the `experiments` folder. 60 | 61 | The hyperparameters for different sites in the paper are set as follows: 62 | 63 | | Hyperparameter | Site A | Site B | Site C | Site D | Site E | Site F | 64 | | --- | --- | --- | --- | --- | --- | --- | 65 | |Dimension of LSTM hidden features | 256 | 512 | 256 | 512 | 256 | 256 | 66 | | Number of LSTM layers | 2 | 2 | 2 | 2 | 2 | 3 | 67 | -------------------------------------------------------------------------------- /config.py: -------------------------------------------------------------------------------- 1 | START_V_I = 1650 # start_vertical_index 2 | START_H_I = 1650 # start_horizontal_index 3 | SIDE_LEN = 1700 # side length 4 | 5 | INTRPL_START_DATE_STR = "0422" # interpolated_start_date 6 | INTRPL_END_DATE_STR = "0923" # interpolated_end_date 7 | 8 | SEED = 313 # random seed 9 | -------------------------------------------------------------------------------- /experiments/spatial_transfer.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": null, 6 | "metadata": { 7 | "ExecuteTime": { 8 | "end_time": "2020-04-19T08:23:25.757521Z", 9 | "start_time": "2020-04-19T08:23:25.753117Z" 10 | } 11 | }, 12 | "outputs": [], 13 | "source": [ 14 | "import os\n", 15 | "import sys\n", 16 | "module_path = os.path.abspath(os.path.join(\"..\"))\n", 17 | "if module_path not in sys.path:\n", 18 | " sys.path.append(module_path)" 19 | ] 20 | }, 21 | { 22 | "cell_type": "code", 23 | "execution_count": null, 24 | "metadata": { 25 | "ExecuteTime": { 26 | "end_time": "2020-04-19T08:23:28.755303Z", 27 | "start_time": "2020-04-19T08:23:25.760710Z" 28 | } 29 | }, 30 | "outputs": [], 31 | "source": [ 32 | "import numpy as np\n", 33 | "import pandas as pd\n", 34 | "from sklearn.metrics import accuracy_score\n", 35 | "import torch\n", 36 | "import torch.nn as nn\n", 37 | "from utils.logger import PrettyLogger\n", 38 | "from utils.io_func import save_to_csv, load_from_pkl, load_from_pth\n", 39 | "from utils.helper import DCMHelper" 40 | ] 41 | }, 42 | { 43 | "cell_type": "code", 44 | "execution_count": null, 45 | "metadata": { 46 | "ExecuteTime": { 47 | "end_time": "2020-04-19T08:23:28.763085Z", 48 | "start_time": "2020-04-19T08:23:28.758935Z" 49 | } 50 | }, 51 | "outputs": [], 52 | "source": [ 53 | "logger = PrettyLogger()\n", 54 | "helper = DCMHelper()" 55 | ] 56 | }, 57 | { 58 | "cell_type": "code", 59 | "execution_count": null, 60 | "metadata": { 61 | "ExecuteTime": { 62 | "end_time": "2020-04-19T08:23:28.775382Z", 63 | "start_time": "2020-04-19T08:23:28.766284Z" 64 | } 65 | }, 66 | "outputs": [], 67 | "source": [ 68 | "BASE_SITES = [\"Site_A\"]\n", 69 | "TEST_SITE = \"Site_B\"\n", 70 | "TEST_YEARS = [str(year) for year in [2018]]\n", 71 | "DATA_DIR = \"../preprocessing/out/{}\".format(TEST_SITE)\n", 72 | "X_PATH_TEMPLATE = os.path.join(DATA_DIR, \"x-{year}.npy\")\n", 73 | "Y_PATH_TEMPLATE = os.path.join(DATA_DIR, \"y-{year}.npy\")\n", 74 | "SCALER_PATH = \"./out/training/{}/scaler.pkl\".format(\"_\".join(BASE_SITES))\n", 75 | "MODEL_PATH = \"./out/training/{}/atbilstm.pth\".format(\"_\".join(BASE_SITES))\n", 76 | "RESULT_DIR = \"./out/spatial_tran/{}/{}\".format(\"_\".join(BASE_SITES), TEST_SITE)\n", 77 | "DEVICE = torch.device(\"cuda:0\")" 78 | ] 79 | }, 80 | { 81 | "cell_type": "markdown", 82 | "metadata": {}, 83 | "source": [ 84 | "# Input" 85 | ] 86 | }, 87 | { 88 | "cell_type": "code", 89 | "execution_count": null, 90 | "metadata": { 91 | "ExecuteTime": { 92 | "end_time": "2020-04-19T08:23:30.864974Z", 93 | "start_time": "2020-04-19T08:23:28.778968Z" 94 | } 95 | }, 96 | "outputs": [], 97 | "source": [ 98 | "def get_paths(path_template, years):\n", 99 | " return [path_template.format(year=year) for year in years]\n", 100 | "\n", 101 | "\n", 102 | "x_test = helper.input_x(get_paths(X_PATH_TEMPLATE, TEST_YEARS))\n", 103 | "y_test = helper.input_y(get_paths(Y_PATH_TEMPLATE, TEST_YEARS))" 104 | ] 105 | }, 106 | { 107 | "cell_type": "markdown", 108 | "metadata": {}, 109 | "source": [ 110 | "# Normalization" 111 | ] 112 | }, 113 | { 114 | "cell_type": "code", 115 | "execution_count": null, 116 | "metadata": { 117 | "ExecuteTime": { 118 | "end_time": "2020-04-19T08:23:31.323735Z", 119 | "start_time": "2020-04-19T08:23:30.867103Z" 120 | } 121 | }, 122 | "outputs": [], 123 | "source": [ 124 | "scaler = load_from_pkl(SCALER_PATH)\n", 125 | "x_test = helper.normalize_with_scaler(scaler, x_test)" 126 | ] 127 | }, 128 | { 129 | "cell_type": "markdown", 130 | "metadata": {}, 131 | "source": [ 132 | "# Prediction" 133 | ] 134 | }, 135 | { 136 | "cell_type": "code", 137 | "execution_count": null, 138 | "metadata": { 139 | "ExecuteTime": { 140 | "end_time": "2020-04-19T08:23:55.970104Z", 141 | "start_time": "2020-04-19T08:23:31.326136Z" 142 | } 143 | }, 144 | "outputs": [], 145 | "source": [ 146 | "test_dataloader = helper.make_data_loader(x_test, y_test, shuffle=False)\n", 147 | "\n", 148 | "net = helper.build_model()\n", 149 | "net.load_state_dict(load_from_pth(MODEL_PATH))\n", 150 | "net = nn.DataParallel(net, device_ids=[0, 1, 2, 3])\n", 151 | "net.to(DEVICE)\n", 152 | "\n", 153 | "y_test_soft_pred, y_test_hard_pred, attn_test = helper.predict(\n", 154 | " net, test_dataloader, DEVICE\n", 155 | ")\n", 156 | "acc_test = accuracy_score(y_test, y_test_hard_pred)\n", 157 | "logger.info(TEST_SITE, \"test acc:\", acc_test)" 158 | ] 159 | }, 160 | { 161 | "cell_type": "markdown", 162 | "metadata": {}, 163 | "source": [ 164 | "# Saving all" 165 | ] 166 | }, 167 | { 168 | "cell_type": "code", 169 | "execution_count": null, 170 | "metadata": { 171 | "ExecuteTime": { 172 | "end_time": "2020-04-19T08:24:03.541550Z", 173 | "start_time": "2020-04-19T08:23:55.972732Z" 174 | } 175 | }, 176 | "outputs": [], 177 | "source": [ 178 | "save_to_csv(\n", 179 | " y_test_soft_pred, os.path.join(RESULT_DIR, \"y_test_soft_pred.csv\")\n", 180 | ")\n", 181 | "save_to_csv(\n", 182 | " y_test_hard_pred, os.path.join(RESULT_DIR, \"y_test_hard_pred.csv\")\n", 183 | ")\n", 184 | "save_to_csv(\n", 185 | " np.array([[acc_test]]),\n", 186 | " os.path.join(RESULT_DIR, \"perf_abstract.csv\"),\n", 187 | " header=[\"acc_test\"]\n", 188 | ")\n", 189 | "save_to_csv(\n", 190 | " helper.test_time_list,\n", 191 | " os.path.join(RESULT_DIR, \"test_time.csv\"),\n", 192 | " header=[\"test_start_time\", \"test_end_time\", \"duration\"]\n", 193 | ")" 194 | ] 195 | } 196 | ], 197 | "metadata": { 198 | "kernelspec": { 199 | "display_name": "Python 3", 200 | "language": "python", 201 | "name": "python3" 202 | }, 203 | "language_info": { 204 | "codemirror_mode": { 205 | "name": "ipython", 206 | "version": 3 207 | }, 208 | "file_extension": ".py", 209 | "mimetype": "text/x-python", 210 | "name": "python", 211 | "nbconvert_exporter": "python", 212 | "pygments_lexer": "ipython3", 213 | "version": "3.5.2" 214 | }, 215 | "toc": { 216 | "base_numbering": 1, 217 | "nav_menu": { 218 | "height": "225px", 219 | "width": "262px" 220 | }, 221 | "number_sections": true, 222 | "sideBar": true, 223 | "skip_h1_title": false, 224 | "title_cell": "Table of Contents", 225 | "title_sidebar": "Contents", 226 | "toc_cell": false, 227 | "toc_position": { 228 | "height": "calc(100% - 180px)", 229 | "left": "10px", 230 | "top": "150px", 231 | "width": "223px" 232 | }, 233 | "toc_section_display": true, 234 | "toc_window_display": true 235 | } 236 | }, 237 | "nbformat": 4, 238 | "nbformat_minor": 2 239 | } 240 | -------------------------------------------------------------------------------- /experiments/training.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": null, 6 | "metadata": { 7 | "ExecuteTime": { 8 | "end_time": "2020-04-19T08:36:08.479621Z", 9 | "start_time": "2020-04-19T08:36:08.475766Z" 10 | } 11 | }, 12 | "outputs": [], 13 | "source": [ 14 | "import os\n", 15 | "import sys\n", 16 | "module_path = os.path.abspath(os.path.join(\"..\"))\n", 17 | "if module_path not in sys.path:\n", 18 | " sys.path.append(module_path)" 19 | ] 20 | }, 21 | { 22 | "cell_type": "code", 23 | "execution_count": null, 24 | "metadata": { 25 | "ExecuteTime": { 26 | "end_time": "2020-04-19T08:36:11.762000Z", 27 | "start_time": "2020-04-19T08:36:08.482310Z" 28 | } 29 | }, 30 | "outputs": [], 31 | "source": [ 32 | "import numpy as np\n", 33 | "import pandas as pd\n", 34 | "from sklearn.metrics import accuracy_score\n", 35 | "import torch\n", 36 | "import torch.nn as nn\n", 37 | "from utils.logger import PrettyLogger\n", 38 | "from utils.io_func import save_to_csv, save_to_pkl, save_to_pth\n", 39 | "from utils.helper import DCMHelper\n", 40 | "from config import SEED" 41 | ] 42 | }, 43 | { 44 | "cell_type": "code", 45 | "execution_count": null, 46 | "metadata": { 47 | "ExecuteTime": { 48 | "end_time": "2020-04-19T08:36:11.768990Z", 49 | "start_time": "2020-04-19T08:36:11.765512Z" 50 | } 51 | }, 52 | "outputs": [], 53 | "source": [ 54 | "logger = PrettyLogger()\n", 55 | "helper = DCMHelper()" 56 | ] 57 | }, 58 | { 59 | "cell_type": "code", 60 | "execution_count": null, 61 | "metadata": { 62 | "ExecuteTime": { 63 | "end_time": "2020-04-19T08:36:11.777101Z", 64 | "start_time": "2020-04-19T08:36:11.771775Z" 65 | } 66 | }, 67 | "outputs": [], 68 | "source": [ 69 | "BASE_SITES = [\"Site_A\"]\n", 70 | "TRAIN_YEARS = [str(year) for year in [2015, 2016, 2017]]\n", 71 | "TEST_YEARS = [str(year) for year in [2018]]\n", 72 | "DATA_DIR_TEMPLATE = \"../preprocessing/out/{site}/\"\n", 73 | "X_PATH_TEMPLATE = os.path.join(DATA_DIR_TEMPLATE, \"x-{year}.npy\")\n", 74 | "Y_PATH_TEMPLATE = os.path.join(DATA_DIR_TEMPLATE, \"y-{year}.npy\")\n", 75 | "RESULT_DIR = \"./out/DCM-20200101/{}/\".format(\"_\".join(BASE_SITES))\n", 76 | "DEVICE = torch.device(\"cuda:0\")" 77 | ] 78 | }, 79 | { 80 | "cell_type": "markdown", 81 | "metadata": {}, 82 | "source": [ 83 | "# Input" 84 | ] 85 | }, 86 | { 87 | "cell_type": "code", 88 | "execution_count": null, 89 | "metadata": { 90 | "ExecuteTime": { 91 | "end_time": "2020-04-19T08:36:32.601886Z", 92 | "start_time": "2020-04-19T08:36:11.780395Z" 93 | } 94 | }, 95 | "outputs": [], 96 | "source": [ 97 | "def get_paths(path_template, sites, years):\n", 98 | " paths = []\n", 99 | " for site in sites:\n", 100 | " for year in years:\n", 101 | " paths.append(path_template.format(site=site, year=year))\n", 102 | " return paths\n", 103 | "\n", 104 | "\n", 105 | "x_train = helper.input_x(get_paths(X_PATH_TEMPLATE, BASE_SITES, TRAIN_YEARS))\n", 106 | "y_train = helper.input_y(get_paths(Y_PATH_TEMPLATE, BASE_SITES, TRAIN_YEARS))\n", 107 | "x_test = helper.input_x(get_paths(X_PATH_TEMPLATE, BASE_SITES, TEST_YEARS))\n", 108 | "y_test = helper.input_y(get_paths(Y_PATH_TEMPLATE, BASE_SITES, TEST_YEARS))" 109 | ] 110 | }, 111 | { 112 | "cell_type": "markdown", 113 | "metadata": {}, 114 | "source": [ 115 | "# Normalization" 116 | ] 117 | }, 118 | { 119 | "cell_type": "code", 120 | "execution_count": null, 121 | "metadata": { 122 | "ExecuteTime": { 123 | "end_time": "2020-04-19T08:36:40.013263Z", 124 | "start_time": "2020-04-19T08:36:32.604725Z" 125 | } 126 | }, 127 | "outputs": [], 128 | "source": [ 129 | "scaler, x_train, x_test = helper.normalize_without_scaler(x_train, x_test)" 130 | ] 131 | }, 132 | { 133 | "cell_type": "markdown", 134 | "metadata": {}, 135 | "source": [ 136 | "# Training models" 137 | ] 138 | }, 139 | { 140 | "cell_type": "code", 141 | "execution_count": null, 142 | "metadata": { 143 | "ExecuteTime": { 144 | "end_time": "2020-04-19T08:36:40.014295Z", 145 | "start_time": "2020-04-19T08:36:44.267Z" 146 | }, 147 | "scrolled": true 148 | }, 149 | "outputs": [], 150 | "source": [ 151 | "train_dataloader = helper.make_data_loader(x_train, y_train, shuffle=True)\n", 152 | "test_dataloader = helper.make_data_loader(x_test, y_test, shuffle=False)\n", 153 | "\n", 154 | "net = helper.build_model()\n", 155 | "helper.init_parameters(net)\n", 156 | "net = nn.DataParallel(net, device_ids=[0, 1, 2, 3])\n", 157 | "net.to(DEVICE)\n", 158 | "\n", 159 | "loss_train_list, acc_train_list, attn_train_list = [], [], []\n", 160 | "loss_test_list, acc_test_list, attn_test_list = [], [], []\n", 161 | "helper.train_model(\n", 162 | " net, train_dataloader, test_dataloader, DEVICE, logger,\n", 163 | " loss_train_list, acc_train_list, attn_train_list,\n", 164 | " loss_test_list, acc_test_list, attn_test_list,\n", 165 | ")" 166 | ] 167 | }, 168 | { 169 | "cell_type": "markdown", 170 | "metadata": {}, 171 | "source": [ 172 | "# Prediction" 173 | ] 174 | }, 175 | { 176 | "cell_type": "code", 177 | "execution_count": null, 178 | "metadata": { 179 | "ExecuteTime": { 180 | "end_time": "2020-04-19T08:36:40.015386Z", 181 | "start_time": "2020-04-19T08:36:44.269Z" 182 | } 183 | }, 184 | "outputs": [], 185 | "source": [ 186 | "y_train_soft_pred, y_train_hard_pred, attn_train = helper.predict(\n", 187 | " net, helper.make_data_loader(x_train, y_train, shuffle=False), DEVICE\n", 188 | ")\n", 189 | "y_test_soft_pred, y_test_hard_pred, attn_test = helper.predict(\n", 190 | " net, test_dataloader, DEVICE\n", 191 | ")\n", 192 | "acc_train = accuracy_score(y_train, y_train_hard_pred)\n", 193 | "acc_test = accuracy_score(y_test, y_test_hard_pred)\n", 194 | "logger.info(\"train acc:\", acc_train, \"test acc:\", acc_test)" 195 | ] 196 | }, 197 | { 198 | "cell_type": "markdown", 199 | "metadata": {}, 200 | "source": [ 201 | "# Saving all" 202 | ] 203 | }, 204 | { 205 | "cell_type": "code", 206 | "execution_count": null, 207 | "metadata": { 208 | "ExecuteTime": { 209 | "end_time": "2020-04-19T08:36:40.016294Z", 210 | "start_time": "2020-04-19T08:36:44.273Z" 211 | } 212 | }, 213 | "outputs": [], 214 | "source": [ 215 | "save_to_csv(\n", 216 | " y_train_soft_pred, os.path.join(RESULT_DIR, \"y_train_soft_pred.csv\")\n", 217 | ")\n", 218 | "save_to_csv(\n", 219 | " y_test_soft_pred, os.path.join(RESULT_DIR, \"y_test_soft_pred.csv\")\n", 220 | ")\n", 221 | "save_to_csv(\n", 222 | " y_train_hard_pred, os.path.join(RESULT_DIR, \"y_train_hard_pred.csv\")\n", 223 | ")\n", 224 | "save_to_csv(\n", 225 | " y_test_hard_pred, os.path.join(RESULT_DIR, \"y_test_hard_pred.csv\")\n", 226 | ")\n", 227 | "save_to_csv(\n", 228 | " np.array([\n", 229 | " loss_train_list, loss_test_list, acc_train_list, acc_test_list\n", 230 | " ]).T,\n", 231 | " os.path.join(RESULT_DIR, \"training_record.csv\"),\n", 232 | " header=[\"training loss\", \"test loss\", \"training acc\", \"test acc\"]\n", 233 | ")\n", 234 | "save_to_csv(\n", 235 | " np.array([[acc_train, acc_test]]),\n", 236 | " os.path.join(RESULT_DIR, \"perf_abstract.csv\"),\n", 237 | " header=[\"acc_train\", \"acc_test\"]\n", 238 | ")\n", 239 | "save_to_pkl(scaler, os.path.join(RESULT_DIR, \"scaler.pkl\"))\n", 240 | "save_to_pth(net, os.path.join(RESULT_DIR, \"atbilstm.pth\"))\n", 241 | "save_to_csv(\n", 242 | " helper.train_time_list,\n", 243 | " os.path.join(RESULT_DIR, \"train_time.csv\"),\n", 244 | " header=[\"train_start_time\", \"train_end_time\", \"duration\"]\n", 245 | ")\n", 246 | "save_to_csv(\n", 247 | " helper.test_time_list,\n", 248 | " os.path.join(RESULT_DIR, \"test_time.csv\"),\n", 249 | " header=[\"test_start_time\", \"test_end_time\", \"duration\"]\n", 250 | ")" 251 | ] 252 | } 253 | ], 254 | "metadata": { 255 | "kernelspec": { 256 | "display_name": "Python 3", 257 | "language": "python", 258 | "name": "python3" 259 | }, 260 | "language_info": { 261 | "codemirror_mode": { 262 | "name": "ipython", 263 | "version": 3 264 | }, 265 | "file_extension": ".py", 266 | "mimetype": "text/x-python", 267 | "name": "python", 268 | "nbconvert_exporter": "python", 269 | "pygments_lexer": "ipython3", 270 | "version": "3.5.2" 271 | }, 272 | "toc": { 273 | "base_numbering": "1", 274 | "nav_menu": { 275 | "height": "225px", 276 | "width": "262px" 277 | }, 278 | "number_sections": true, 279 | "sideBar": true, 280 | "skip_h1_title": false, 281 | "title_cell": "Table of Contents", 282 | "title_sidebar": "Contents", 283 | "toc_cell": false, 284 | "toc_position": { 285 | "height": "calc(100% - 180px)", 286 | "left": "10px", 287 | "top": "150px", 288 | "width": "189px" 289 | }, 290 | "toc_section_display": true, 291 | "toc_window_display": true 292 | } 293 | }, 294 | "nbformat": 4, 295 | "nbformat_minor": 2 296 | } 297 | -------------------------------------------------------------------------------- /models/dcm.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import torch 3 | import torch.nn as nn 4 | import torch.nn.functional as F 5 | 6 | 7 | class DCM(nn.Module): 8 | def __init__( 9 | self, seed, input_feature_size, hidden_size, num_layers, 10 | bidirectional, dropout, num_classes 11 | ): 12 | super().__init__() 13 | self._set_reproducible(seed) 14 | 15 | self.lstm = nn.LSTM( 16 | input_size=input_feature_size, 17 | hidden_size=hidden_size, 18 | num_layers=num_layers, 19 | bidirectional=bidirectional, 20 | batch_first=True, 21 | dropout=dropout, 22 | ) # i/o: (batch, seq_len, num_directions*input_/hidden_size) 23 | num_directions = 2 if bidirectional else 1 24 | self.attention = nn.Linear( 25 | in_features=num_directions * hidden_size, 26 | out_features=1, 27 | ) 28 | self.fc = nn.Linear( 29 | in_features=num_directions * hidden_size, 30 | out_features=num_classes, 31 | ) 32 | 33 | def _set_reproducible(self, seed, cudnn=False): 34 | np.random.seed(seed) 35 | torch.manual_seed(seed) 36 | if cudnn: 37 | torch.backends.cudnn.deterministic = True 38 | torch.backends.cudnn.benchmark = False 39 | 40 | def forward(self, x): 41 | self.lstm.flatten_parameters() 42 | # lstm_out: (batch, seq_len, num_directions*hidden_size) 43 | lstm_out, _ = self.lstm(x) 44 | # softmax along seq_len axis 45 | attn_weights = F.softmax(F.relu(self.attention(lstm_out)), dim=1) 46 | # attn (after permutation): (batch, 1, seq_len) 47 | fc_in = attn_weights.permute(0, 2, 1).bmm(lstm_out) 48 | fc_out = self.fc(fc_in) 49 | return fc_out.squeeze(), attn_weights.squeeze() -------------------------------------------------------------------------------- /preprocessing/preprocess_ARD.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": null, 6 | "metadata": { 7 | "ExecuteTime": { 8 | "end_time": "2020-01-09T01:19:39.422963Z", 9 | "start_time": "2020-01-09T01:19:39.417422Z" 10 | } 11 | }, 12 | "outputs": [], 13 | "source": [ 14 | "import os\n", 15 | "import sys\n", 16 | "module_path = os.path.abspath(os.path.join(\"..\"))\n", 17 | "if module_path not in sys.path:\n", 18 | " sys.path.append(module_path)" 19 | ] 20 | }, 21 | { 22 | "cell_type": "code", 23 | "execution_count": null, 24 | "metadata": { 25 | "ExecuteTime": { 26 | "end_time": "2020-01-09T01:19:59.036142Z", 27 | "start_time": "2020-01-09T01:19:39.435378Z" 28 | } 29 | }, 30 | "outputs": [], 31 | "source": [ 32 | "import numpy as np\n", 33 | "from utils.logger import PrettyLogger\n", 34 | "from utils.date import str2date, int2date_delta, date2str\n", 35 | "from utils.io_func import save_to_npy, load_from_tiff\n", 36 | "from config import (\n", 37 | " START_V_I, START_H_I, SIDE_LEN, INTRPL_START_DATE_STR, INTRPL_END_DATE_STR\n", 38 | ")" 39 | ] 40 | }, 41 | { 42 | "cell_type": "code", 43 | "execution_count": null, 44 | "metadata": { 45 | "ExecuteTime": { 46 | "end_time": "2020-01-09T01:19:59.043143Z", 47 | "start_time": "2020-01-09T01:19:59.039422Z" 48 | } 49 | }, 50 | "outputs": [], 51 | "source": [ 52 | "logger = PrettyLogger()" 53 | ] 54 | }, 55 | { 56 | "cell_type": "code", 57 | "execution_count": null, 58 | "metadata": { 59 | "ExecuteTime": { 60 | "end_time": "2020-01-09T01:19:59.055840Z", 61 | "start_time": "2020-01-09T01:19:59.045850Z" 62 | } 63 | }, 64 | "outputs": [], 65 | "source": [ 66 | "SITE = \"Site_A\"\n", 67 | "YEAR = \"2015\"\n", 68 | "DATA_DIR = \"../data/{}/ARD/{}/\".format(SITE, YEAR)\n", 69 | "OUTPUT_DIR = \"./out/{}/ARD/cropped_interpolated/{}/\".format(SITE, YEAR)\n", 70 | "AVAI_PATH = os.path.join(OUTPUT_DIR, \"availability.npy\")\n", 71 | "FILTER_BAND_PATH = os.path.join(OUTPUT_DIR, \"filter_band.npy\")\n", 72 | "INTERPOLATED_PATH = os.path.join(OUTPUT_DIR, \"interpolated.npy\")\n", 73 | "FINAL_OUTOUT_FILEPATH = \"./out/{}/x-{}.npy\".format(SITE, YEAR)" 74 | ] 75 | }, 76 | { 77 | "cell_type": "markdown", 78 | "metadata": {}, 79 | "source": [ 80 | "# ARD observation input, cropping and filtering" 81 | ] 82 | }, 83 | { 84 | "cell_type": "code", 85 | "execution_count": null, 86 | "metadata": { 87 | "ExecuteTime": { 88 | "end_time": "2020-01-09T01:19:59.109273Z", 89 | "start_time": "2020-01-09T01:19:59.059000Z" 90 | } 91 | }, 92 | "outputs": [], 93 | "source": [ 94 | "# link the filenames to date\n", 95 | "date_filename_dict = {}\n", 96 | "for filename in sorted(os.listdir(DATA_DIR)):\n", 97 | " date = str2date(filename[15:23])\n", 98 | " if (\n", 99 | " date >= str2date(\"{}{}\".format(YEAR, INTRPL_START_DATE_STR))\n", 100 | " and date <= str2date(\"{}{}\".format(YEAR, INTRPL_END_DATE_STR))\n", 101 | " ):\n", 102 | " if date not in date_filename_dict.keys():\n", 103 | " date_filename_dict[date] = []\n", 104 | " date_filename_dict[date].append(filename)" 105 | ] 106 | }, 107 | { 108 | "cell_type": "code", 109 | "execution_count": null, 110 | "metadata": { 111 | "ExecuteTime": { 112 | "end_time": "2020-01-09T01:36:33.301768Z", 113 | "start_time": "2020-01-09T01:19:59.112110Z" 114 | }, 115 | "scrolled": true 116 | }, 117 | "outputs": [], 118 | "source": [ 119 | "# read ARD images, crop ARD images and detect invalid values\n", 120 | "raw_dates = sorted(date_filename_dict.keys())\n", 121 | "availability = np.zeros((SIDE_LEN, SIDE_LEN, len(raw_dates)))\n", 122 | "valid = np.zeros((SIDE_LEN, SIDE_LEN, len(raw_dates), 6))\n", 123 | "\n", 124 | "to_fill = np.vectorize(lambda x: int(\"{:011b}\".format(x)[-1], 2))\n", 125 | "to_clear = np.vectorize(lambda x: int(\"{:011b}\".format(x)[-2], 2))\n", 126 | "to_cloud_shadow = np.vectorize(lambda x: int(\"{:011b}\".format(x)[-4], 2))\n", 127 | "to_cloud = np.vectorize(lambda x: int(\"{:011b}\".format(x)[-6], 2))\n", 128 | "for i, date in enumerate(raw_dates):\n", 129 | " logger.info(\"Loading: {}/{}\".format(i+1, len(raw_dates)), date2str(date))\n", 130 | " sr_bands = []\n", 131 | " for filename in date_filename_dict[date]:\n", 132 | " band = load_from_tiff(os.path.join(DATA_DIR, filename))[\n", 133 | " START_V_I:START_V_I+SIDE_LEN, START_H_I:START_H_I+SIDE_LEN\n", 134 | " ]\n", 135 | " if filename[-11:-4] != \"PIXELQA\":\n", 136 | " sr_bands.append(band)\n", 137 | " else:\n", 138 | " qa_band = band\n", 139 | " sr_bands = np.array(sr_bands).transpose((1, 2, 0))\n", 140 | "\n", 141 | " flag_sr_range = ((sr_bands >= 0) & (sr_bands <= 10000)).all(axis=2)\n", 142 | " fill_band = to_fill(qa_band)\n", 143 | " flag_fill = (fill_band == 0)\n", 144 | " clear_band = to_clear(qa_band)\n", 145 | " flag_clear = (clear_band == 1)\n", 146 | " cloud_shadow_band = to_cloud_shadow(qa_band)\n", 147 | " flag_cloud_shadow = (cloud_shadow_band == 0)\n", 148 | " cloud_band = to_cloud(qa_band)\n", 149 | " flag_cloud = (cloud_band == 0)\n", 150 | " flag = flag_sr_range*flag_fill*flag_clear*flag_cloud_shadow*flag_cloud\n", 151 | "\n", 152 | " availability[:, :, i] = flag\n", 153 | "\n", 154 | " # make invalid observations zero, only for the convenience of debugging\n", 155 | " valid[:, :, i, :] = sr_bands\n", 156 | " valid[:, :, i, :] = valid[:, :, i, :]*(flag.reshape(*flag.shape, 1))\n", 157 | "\n", 158 | "save_to_npy(availability, AVAI_PATH)" 159 | ] 160 | }, 161 | { 162 | "cell_type": "code", 163 | "execution_count": null, 164 | "metadata": { 165 | "ExecuteTime": { 166 | "end_time": "2020-01-09T01:36:34.384851Z", 167 | "start_time": "2020-01-09T01:36:33.306697Z" 168 | } 169 | }, 170 | "outputs": [], 171 | "source": [ 172 | "\"\"\"\n", 173 | "========== PIXEL FILTER METHOD BY AVAILABILITY ==========\n", 174 | "If the number of valid observations after May 15 >= 7,\n", 175 | "the pixel will be included in the dataset, otherwise it will be excluded.\n", 176 | "\"\"\"\n", 177 | "\n", 178 | "index4filter = raw_dates.index(list(filter(\n", 179 | " lambda x: x > str2date(\"{}0515\".format(YEAR)), raw_dates\n", 180 | "))[0])\n", 181 | "filter_band = availability[:, :, index4filter:].sum(axis=2) >= 7\n", 182 | "logger.info(\"Validity percentage ({} {}): {:.4f}\".format(\n", 183 | " SITE, YEAR,\n", 184 | " filter_band.sum()/(filter_band.shape[0]*filter_band.shape[1])\n", 185 | "))\n", 186 | "save_to_npy(filter_band, FILTER_BAND_PATH)" 187 | ] 188 | }, 189 | { 190 | "cell_type": "markdown", 191 | "metadata": {}, 192 | "source": [ 193 | "# Temporal interpolation" 194 | ] 195 | }, 196 | { 197 | "cell_type": "code", 198 | "execution_count": null, 199 | "metadata": { 200 | "ExecuteTime": { 201 | "end_time": "2020-01-09T01:36:38.032317Z", 202 | "start_time": "2020-01-09T01:36:36.930790Z" 203 | } 204 | }, 205 | "outputs": [], 206 | "source": [ 207 | "# prepare target dates for interpolation\n", 208 | "intrpl_start_date = str2date(\"{}{}\".format(YEAR, INTRPL_START_DATE_STR))\n", 209 | "intrpl_end_date = str2date(\"{}{}\".format(YEAR, INTRPL_END_DATE_STR))\n", 210 | "intrpl_delta_days = list(range(\n", 211 | " 0, (intrpl_end_date - intrpl_start_date).days + 1, 7\n", 212 | "))\n", 213 | "intrpl_dates = [\n", 214 | " int2date_delta(intrpl_delta_day) + intrpl_start_date\n", 215 | " for intrpl_delta_day in intrpl_delta_days\n", 216 | "]" 217 | ] 218 | }, 219 | { 220 | "cell_type": "code", 221 | "execution_count": null, 222 | "metadata": { 223 | "ExecuteTime": { 224 | "end_time": "2020-01-09T03:00:21.154986Z", 225 | "start_time": "2020-01-09T01:36:38.035281Z" 226 | }, 227 | "scrolled": true 228 | }, 229 | "outputs": [], 230 | "source": [ 231 | "'''\n", 232 | "========== INTERPOLATION METHOD ==========\n", 233 | "situation I (normal): d_1, d_2*, target, d_3*, d_4, ...\n", 234 | "situation II (close to the start date): target, d_1*, d_2*, d_3, ...\n", 235 | "situation III (close to the end date): d_1, d_2, ..., d_(-2), d_(-1), target\n", 236 | "'''\n", 237 | "\n", 238 | "interpolated = np.zeros((SIDE_LEN, SIDE_LEN, len(intrpl_dates), 6))\n", 239 | "for intrpl_date_index, intrpl_date in enumerate(intrpl_dates):\n", 240 | " logger.info(\"Interpolating: {}/{} {} \".format(\n", 241 | " intrpl_date_index + 1, len(intrpl_dates), date2str(intrpl_date))\n", 242 | " )\n", 243 | " # descending/ascending order for searching the nearest day before/after\n", 244 | " before_dates = list(filter(lambda x: x <= intrpl_date, raw_dates))[::-1]\n", 245 | " after_dates = list(filter(lambda x: x >= intrpl_date, raw_dates))\n", 246 | "\n", 247 | " for i in range(SIDE_LEN):\n", 248 | " for j in range(SIDE_LEN):\n", 249 | "\n", 250 | " # filter invalid pixel\n", 251 | " if not filter_band[i, j]:\n", 252 | " continue\n", 253 | "\n", 254 | " # situation I\n", 255 | " d_1 = None\n", 256 | " for nearest_before_index, before_date in enumerate(before_dates):\n", 257 | " before_date_raw_index = raw_dates.index(before_date)\n", 258 | " if availability[i, j][before_date_raw_index]:\n", 259 | " d_1 = before_date\n", 260 | " date_raw_index_1 = before_date_raw_index\n", 261 | " break\n", 262 | " d_2 = None\n", 263 | " for nearest_after_index, after_date in enumerate(after_dates):\n", 264 | " after_date_raw_index = raw_dates.index(after_date)\n", 265 | " if availability[i, j][after_date_raw_index]:\n", 266 | " d_2 = after_date\n", 267 | " date_raw_index_2 = after_date_raw_index\n", 268 | " break\n", 269 | "\n", 270 | " # situation II: search the second nearest after date\n", 271 | " if not d_1:\n", 272 | " for after_date in after_dates[nearest_after_index+1:]:\n", 273 | " after_date_raw_index = raw_dates.index(after_date)\n", 274 | " if availability[i, j][after_date_raw_index]:\n", 275 | " d_1 = after_date\n", 276 | " date_raw_index_1 = after_date_raw_index\n", 277 | " break\n", 278 | "\n", 279 | " # situation III: search the second nearest before date\n", 280 | " if not d_2:\n", 281 | " for before_date in before_dates[nearest_before_index+1:]:\n", 282 | " before_date_raw_index = raw_dates.index(before_date)\n", 283 | " if availability[i, j][before_date_raw_index]:\n", 284 | " d_2 = before_date\n", 285 | " date_raw_index_2 = before_date_raw_index\n", 286 | " break\n", 287 | "\n", 288 | " interpolated[i][j][intrpl_date_index] = [np.interp(\n", 289 | " (intrpl_date-d_1).days,\n", 290 | " [0, (d_2-d_1).days],\n", 291 | " [valid[i, j, date_raw_index_1, band_index],\n", 292 | " valid[i, j, date_raw_index_2, band_index]]\n", 293 | " ) for band_index in range(6)]\n", 294 | "\n", 295 | "save_to_npy(interpolated, INTERPOLATED_PATH)\n", 296 | "x = interpolated[filter_band]\n", 297 | "save_to_npy(x, FINAL_OUTOUT_FILEPATH)" 298 | ] 299 | } 300 | ], 301 | "metadata": { 302 | "kernelspec": { 303 | "display_name": "Python 3", 304 | "language": "python", 305 | "name": "python3" 306 | }, 307 | "language_info": { 308 | "codemirror_mode": { 309 | "name": "ipython", 310 | "version": 3 311 | }, 312 | "file_extension": ".py", 313 | "mimetype": "text/x-python", 314 | "name": "python", 315 | "nbconvert_exporter": "python", 316 | "pygments_lexer": "ipython3", 317 | "version": "3.5.2" 318 | }, 319 | "toc": { 320 | "base_numbering": 1, 321 | "nav_menu": {}, 322 | "number_sections": true, 323 | "sideBar": true, 324 | "skip_h1_title": false, 325 | "title_cell": "Table of Contents", 326 | "title_sidebar": "Contents", 327 | "toc_cell": false, 328 | "toc_position": { 329 | "height": "calc(100% - 180px)", 330 | "left": "10px", 331 | "top": "150px", 332 | "width": "235px" 333 | }, 334 | "toc_section_display": true, 335 | "toc_window_display": true 336 | } 337 | }, 338 | "nbformat": 4, 339 | "nbformat_minor": 2 340 | } 341 | -------------------------------------------------------------------------------- /preprocessing/preprocess_CDL.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": null, 6 | "metadata": { 7 | "ExecuteTime": { 8 | "end_time": "2020-01-09T06:35:28.038350Z", 9 | "start_time": "2020-01-09T06:35:28.031637Z" 10 | } 11 | }, 12 | "outputs": [], 13 | "source": [ 14 | "import os\n", 15 | "import sys\n", 16 | "module_path = os.path.abspath(os.path.join(\"..\"))\n", 17 | "if module_path not in sys.path:\n", 18 | " sys.path.append(module_path)" 19 | ] 20 | }, 21 | { 22 | "cell_type": "code", 23 | "execution_count": null, 24 | "metadata": { 25 | "ExecuteTime": { 26 | "end_time": "2020-01-09T06:35:35.435742Z", 27 | "start_time": "2020-01-09T06:35:28.052269Z" 28 | } 29 | }, 30 | "outputs": [], 31 | "source": [ 32 | "import numpy as np\n", 33 | "from utils.logger import PrettyLogger\n", 34 | "from utils.io_func import (\n", 35 | " load_from_tiff, save_to_tiff, load_from_npy, save_to_npy\n", 36 | ")\n", 37 | "from config import START_V_I, START_H_I, SIDE_LEN" 38 | ] 39 | }, 40 | { 41 | "cell_type": "code", 42 | "execution_count": null, 43 | "metadata": { 44 | "ExecuteTime": { 45 | "end_time": "2020-01-09T06:35:35.443189Z", 46 | "start_time": "2020-01-09T06:35:35.439382Z" 47 | } 48 | }, 49 | "outputs": [], 50 | "source": [ 51 | "logger = PrettyLogger()" 52 | ] 53 | }, 54 | { 55 | "cell_type": "code", 56 | "execution_count": null, 57 | "metadata": { 58 | "ExecuteTime": { 59 | "end_time": "2020-01-09T06:35:35.464212Z", 60 | "start_time": "2020-01-09T06:35:35.445754Z" 61 | } 62 | }, 63 | "outputs": [], 64 | "source": [ 65 | "SITE_YEAR_TUPLE = ((\"Site_A\", \"2015\"),)\n", 66 | "DATA_DIR_TEMPLATE = \"../data/{site}/CDL/\"\n", 67 | "CROPPED_FILEPATH_TEMPLATE = \"./out/{site}/CDL/cropped/CDL-{year}.tif\"\n", 68 | "TRANSCODED_FILEPATH_TEMPLATE = \"./out/{site}/CDL/transcoded/CDL-{year}.tif\"\n", 69 | "INTERPOLATED_FILEPATH_TEMPLATE = (\n", 70 | " \"./out/{site}/ARD/cropped_interpolated/{year}/filter_band.npy\"\n", 71 | ")\n", 72 | "FINAL_OUT_FILEPATH_TEMPLATE = \"./out/{site}/y-{year}.npy\"" 73 | ] 74 | }, 75 | { 76 | "cell_type": "markdown", 77 | "metadata": {}, 78 | "source": [ 79 | "# Cropping and transcoding" 80 | ] 81 | }, 82 | { 83 | "cell_type": "code", 84 | "execution_count": null, 85 | "metadata": { 86 | "ExecuteTime": { 87 | "end_time": "2020-01-09T06:35:35.493707Z", 88 | "start_time": "2020-01-09T06:35:35.467584Z" 89 | } 90 | }, 91 | "outputs": [], 92 | "source": [ 93 | "def transcode(p):\n", 94 | " # corn:1 -> 1, soybean:5 -> 2, other:other -> 0\n", 95 | " return 1 if p == 1 else 2 if p == 5 else 0\n", 96 | "\n", 97 | "\n", 98 | "def preprocess_cdl(\n", 99 | " data_path, cropped_filepath, transcoded_filepath,\n", 100 | " interpolated_filepath, final_out_filepath\n", 101 | "):\n", 102 | " # input\n", 103 | " raw_img = load_from_tiff(data_path)\n", 104 | "\n", 105 | " # crop cdl\n", 106 | " cropped_img = raw_img[\n", 107 | " START_V_I:START_V_I + SIDE_LEN, START_H_I:START_H_I+SIDE_LEN\n", 108 | " ]\n", 109 | " save_to_tiff(cropped_img, cropped_filepath)\n", 110 | "\n", 111 | " # transcoding\n", 112 | " transcoded_img = np.vectorize(transcode)(cropped_img)\n", 113 | " save_to_tiff(transcoded_img, transcoded_filepath)\n", 114 | "\n", 115 | " # remove invalid pixels\n", 116 | " filter_band = load_from_npy(interpolated_filepath)\n", 117 | " y = transcoded_img[filter_band]\n", 118 | "\n", 119 | " # output preprocessed data\n", 120 | " save_to_npy(y, final_out_filepath)" 121 | ] 122 | }, 123 | { 124 | "cell_type": "code", 125 | "execution_count": null, 126 | "metadata": { 127 | "ExecuteTime": { 128 | "end_time": "2020-01-09T06:36:15.105776Z", 129 | "start_time": "2020-01-09T06:35:35.496690Z" 130 | } 131 | }, 132 | "outputs": [], 133 | "source": [ 134 | "for site, year in SITE_YEAR_TUPLE:\n", 135 | " data_dir = DATA_DIR_TEMPLATE.format(site=site)\n", 136 | " filename = list(filter(lambda x: x[4:8] == year, os.listdir(data_dir)))[0]\n", 137 | " data_path = os.path.join(data_dir, filename)\n", 138 | " logger.info(\"Processing:\" site, filename)\n", 139 | " preprocess_cdl(\n", 140 | " data_path=data_path,\n", 141 | " cropped_filepath=CROPPED_FILEPATH_TEMPLATE.format(\n", 142 | " site=site, year=year\n", 143 | " ),\n", 144 | " transcoded_filepath=TRANSCODED_FILEPATH_TEMPLATE.format(\n", 145 | " site=site, year=year\n", 146 | " ),\n", 147 | " interpolated_filepath=INTERPOLATED_FILEPATH_TEMPLATE.format(\n", 148 | " site=site, year=year\n", 149 | " ),\n", 150 | " final_out_filepath=FINAL_OUT_FILEPATH_TEMPLATE.format(\n", 151 | " site=site, year=year\n", 152 | " )\n", 153 | " )" 154 | ] 155 | } 156 | ], 157 | "metadata": { 158 | "kernelspec": { 159 | "display_name": "Python 3", 160 | "language": "python", 161 | "name": "python3" 162 | }, 163 | "language_info": { 164 | "codemirror_mode": { 165 | "name": "ipython", 166 | "version": 3 167 | }, 168 | "file_extension": ".py", 169 | "mimetype": "text/x-python", 170 | "name": "python", 171 | "nbconvert_exporter": "python", 172 | "pygments_lexer": "ipython3", 173 | "version": "3.5.2" 174 | }, 175 | "toc": { 176 | "base_numbering": 1, 177 | "nav_menu": {}, 178 | "number_sections": true, 179 | "sideBar": true, 180 | "skip_h1_title": false, 181 | "title_cell": "Table of Contents", 182 | "title_sidebar": "Contents", 183 | "toc_cell": false, 184 | "toc_position": { 185 | "height": "calc(100% - 180px)", 186 | "left": "10px", 187 | "top": "150px", 188 | "width": "231px" 189 | }, 190 | "toc_section_display": true, 191 | "toc_window_display": true 192 | } 193 | }, 194 | "nbformat": 4, 195 | "nbformat_minor": 2 196 | } 197 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | torch 2 | numpy 3 | pandas 4 | scikit-learn 5 | jupyter 6 | -------------------------------------------------------------------------------- /utils/date.py: -------------------------------------------------------------------------------- 1 | import datetime 2 | 3 | 4 | def str2date(date_str, format="%Y%m%d"): 5 | return datetime.datetime.strptime(date_str, format) 6 | 7 | 8 | def date2str(date, format="%Y%m%d"): 9 | return date.strftime(format) 10 | 11 | 12 | def int2date_delta(date_delta_int): 13 | return datetime.timedelta(date_delta_int) -------------------------------------------------------------------------------- /utils/helper.py: -------------------------------------------------------------------------------- 1 | import os 2 | import sys 3 | module_path = os.path.abspath(os.path.join("..")) 4 | if module_path not in sys.path: 5 | sys.path.append(module_path) 6 | 7 | import re 8 | import numpy as np 9 | import torch 10 | from torch.utils.data import Dataset, DataLoader 11 | import torch.nn as nn 12 | import torch.nn.functional as F 13 | import torch.optim as optim 14 | from models.dcm import DCM 15 | from config import SEED 16 | from utils.timer import record_time 17 | 18 | 19 | class DCMHelper(object): 20 | def __init__( 21 | self, seed=SEED, hidden_size=256, num_layers=2, 22 | bidirectional=True, dropout=0.5, batch_size=int(pow(2, 15)), 23 | num_workers=2, drop_last=False, criterion=nn.CrossEntropyLoss(), 24 | max_epoch=300, min_stop_epoch=5, stop_threshold=0.0002, 25 | record_attn=False, lr=0.0005, weight_decay=0.001, use_scheduler=True, 26 | mode="min", factor=0.5, patience=3, threshold=0.002, 27 | threshold_mode="abs", verbose=True 28 | ): 29 | # data description 30 | self.input_feature_size = None 31 | self.num_classes = None 32 | 33 | # model initialization (structure and hyperparameters) 34 | self.seed = seed 35 | self.hidden_size = hidden_size 36 | self.num_layers = num_layers 37 | self.bidirectional = bidirectional 38 | self.num_directions = 2 if self.bidirectional else 1 39 | self.dropout = dropout 40 | 41 | # data loader 42 | self.batch_size = batch_size 43 | self.num_workers = num_workers 44 | self.drop_last = drop_last 45 | 46 | # training process 47 | self.criterion = criterion 48 | self.max_epoch = max_epoch 49 | self.min_stop_epoch = min_stop_epoch 50 | self.stop_threshold = stop_threshold 51 | self.record_attn = record_attn 52 | 53 | # optimizer 54 | self.lr = lr 55 | self.weight_decay = weight_decay 56 | 57 | # scheduler 58 | self.use_scheduler = use_scheduler 59 | self.mode = mode 60 | self.factor = factor 61 | self.patience = patience 62 | self.threshold = threshold 63 | self.threshold_mode = threshold_mode 64 | self.verbose = verbose 65 | 66 | self.train_time_list = [] 67 | self.test_time_list = [] 68 | 69 | def _input_data(self, paths): 70 | data = np.concatenate( 71 | [np.load(path) for path in paths], axis=0 72 | ) 73 | return data 74 | 75 | def input_x(self, paths): 76 | # x: (num_samples, seq_len, input_feature_size) 77 | x = self._input_data(paths).astype("float32") 78 | self.input_feature_size = x.shape[2] 79 | return x 80 | 81 | def input_y(self, paths): 82 | y = self._input_data(paths).astype("int64") 83 | self.num_classes = np.unique(y).shape[0] 84 | return y 85 | 86 | def normalize_with_scaler(self, scaler, x_test): 87 | x_test = scaler( 88 | torch.FloatTensor(x_test.transpose((0, 2, 1))) 89 | ).numpy().transpose((0, 2, 1)) 90 | return x_test 91 | 92 | def normalize_without_scaler(self, x_train, x_test): 93 | scaler = torch.nn.BatchNorm1d( 94 | self.input_feature_size, eps=0, momentum=1, affine=False 95 | ) 96 | scaler.train() 97 | x_train = scaler( 98 | torch.FloatTensor(x_train.transpose((0, 2, 1))) 99 | ).numpy().transpose((0, 2, 1)) 100 | scaler.eval() 101 | x_test = scaler( 102 | torch.FloatTensor(x_test.transpose((0, 2, 1))) 103 | ).numpy().transpose((0, 2, 1)) 104 | return scaler, x_train, x_test 105 | 106 | def _collate_fn(self, batch): 107 | """ 108 | define how to aggregate samples to batch 109 | """ 110 | return { 111 | "x": torch.FloatTensor( 112 | np.array([sample["sample_x"] for sample in batch]) 113 | ), 114 | "y": torch.LongTensor( 115 | np.array([sample["sample_y"] for sample in batch]) 116 | ) 117 | } 118 | 119 | def make_data_loader(self, x, y, shuffle): 120 | return DataLoader( 121 | CropMappingDataset(x, y), 122 | batch_size=self.batch_size, shuffle=shuffle, 123 | num_workers=self.num_workers, collate_fn=self._collate_fn, 124 | drop_last=self.drop_last 125 | ) 126 | 127 | def build_model(self): 128 | return DCM( 129 | seed=self.seed, 130 | input_feature_size=self.input_feature_size, 131 | hidden_size=self.hidden_size, 132 | num_layers=self.num_layers, 133 | bidirectional=self.bidirectional, 134 | dropout=self.dropout, 135 | num_classes=self.num_classes 136 | ) 137 | 138 | def _init_parameters(self, submodule): 139 | if type(submodule) == nn.LSTM: 140 | for name, param in submodule.named_parameters(): 141 | if re.search("bias_ih", name): 142 | # set forget gate bias to 3.0 143 | param.detach().chunk(4)[1].fill_(3.0) 144 | 145 | def init_parameters(self, net): 146 | net.apply(self._init_parameters) 147 | 148 | def _eval_perf(self, net, dataloader, device): 149 | net.eval() 150 | with torch.no_grad(): 151 | attn_batch_list = [] 152 | losses = 0 153 | correct = 0 154 | for i, batch in enumerate(dataloader): 155 | xt, yt = batch["x"].to(device), batch["y"].to(device) 156 | outputs, attn_batch = net(xt) 157 | loss = self.criterion(outputs, yt) 158 | losses += loss.item() * yt.shape[0] 159 | yt_pred = torch.max(outputs, dim=1)[1] 160 | correct += (yt_pred == yt).sum().item() 161 | attn_batch_list.append(attn_batch) 162 | running_loss = losses / len(dataloader.dataset) 163 | acc = correct / len(dataloader.dataset) 164 | attn = torch.cat(attn_batch_list, dim=0).cpu().numpy() 165 | net.train() 166 | return running_loss, acc, attn 167 | 168 | def _train_model( 169 | self, net, train_dataloader, test_dataloader, device, logger, 170 | loss_train_list, acc_train_list, attn_train_list, loss_test_list, 171 | acc_test_list, attn_test_list 172 | ): 173 | optimizer = optim.Adam( 174 | net.parameters(), lr=self.lr, weight_decay=self.weight_decay, 175 | ) 176 | scheduler = optim.lr_scheduler.ReduceLROnPlateau( 177 | optimizer, mode=self.mode, factor=self.factor, 178 | patience=self.patience, threshold=self.threshold, 179 | threshold_mode=self.threshold_mode, verbose=self.verbose 180 | ) 181 | 182 | for epoch in range(1, self.max_epoch + 1): 183 | net.train() 184 | for i, batch in enumerate(train_dataloader): 185 | xt_train_batch = batch["x"].to(device) 186 | yt_train_batch = batch["y"].to(device) 187 | optimizer.zero_grad() 188 | outputs, _ = net(xt_train_batch) 189 | loss = self.criterion(outputs, yt_train_batch) 190 | loss.backward() 191 | optimizer.step() 192 | loss_train, acc_train, attn_train = self._eval_perf( 193 | net, train_dataloader, device 194 | ) 195 | if self.use_scheduler: 196 | scheduler.step(loss_train) # adjust learning rate 197 | loss_test, acc_test, attn_test = self._eval_perf( 198 | net, test_dataloader, device 199 | ) 200 | loss_train_list.append(loss_train) 201 | acc_train_list.append(acc_train) 202 | loss_test_list.append(loss_test) 203 | acc_test_list.append(acc_test) 204 | if self.record_attn: 205 | attn_train_list.append(attn_train) 206 | attn_test_list.append(attn_test) 207 | 208 | logger.info(( 209 | "[epoch {:d}] " 210 | "training loss: {:.4f}, test loss: {:.4f}, " 211 | "training acc: {:.4f}, test acc: {:.4f} " 212 | " (lr => {:f})" 213 | ).format( 214 | epoch, 215 | loss_train, loss_test, 216 | acc_train, acc_test, 217 | optimizer.param_groups[0]["lr"]) 218 | ) 219 | 220 | if ( 221 | epoch >= self.min_stop_epoch 222 | and ( 223 | np.array(acc_train_list[epoch - 5:epoch]).ptp() 224 | <= self.stop_threshold 225 | ) 226 | ): 227 | break 228 | 229 | logger.info("Training completed") 230 | 231 | def train_model( 232 | self, net, train_dataloader, test_dataloader, device, logger, 233 | loss_train_list, acc_train_list, attn_train_list, loss_test_list, 234 | acc_test_list, attn_test_list 235 | ): 236 | record_time(self.train_time_list, self._train_model, [ 237 | net, train_dataloader, test_dataloader, device, logger, 238 | loss_train_list, acc_train_list, attn_train_list, loss_test_list, 239 | acc_test_list, attn_test_list 240 | ]) 241 | 242 | def _predict(self, net, dataloader, device): 243 | yt_soft_pred_batch_list = [] 244 | attn_batch_list = [] 245 | net.eval() 246 | with torch.no_grad(): 247 | for i, batch in enumerate(dataloader): 248 | xt, yt = batch["x"].to(device), batch["y"].to(device) 249 | outputs, attn_batch = net(xt) 250 | yt_soft_pred_batch_list.append(F.softmax(outputs, dim=1)) 251 | attn_batch_list.append(attn_batch) 252 | y_soft_pred = torch.cat( 253 | yt_soft_pred_batch_list, dim=0 254 | ).cpu().numpy() 255 | y_hard_pred = np.argmax(y_soft_pred, axis=1) 256 | attn = torch.cat(attn_batch_list, dim=0).cpu().numpy() 257 | return y_soft_pred, y_hard_pred, attn 258 | 259 | def predict(self, net, dataloader, device): 260 | return record_time(self.test_time_list, self._predict, [ 261 | net, dataloader, device 262 | ]) 263 | 264 | 265 | class CropMappingDataset(Dataset): 266 | """ 267 | crop classification dataset 268 | """ 269 | 270 | def __init__(self, x, y): 271 | self.x = x 272 | self.y = y 273 | 274 | def __len__(self): 275 | return self.x.shape[0] 276 | 277 | def __getitem__(self, idx): 278 | return {"sample_x": self.x[idx], "sample_y": self.y[idx]} 279 | -------------------------------------------------------------------------------- /utils/io_func.py: -------------------------------------------------------------------------------- 1 | import os 2 | import re 3 | import pickle 4 | import numpy as np 5 | import pandas as pd 6 | import matplotlib.pyplot as plt 7 | import tifffile as tiff 8 | import torch 9 | 10 | 11 | def _assert_suffix_match(suffix, path): 12 | assert re.search(r"\.{}$".format(suffix), path), "suffix mismatch" 13 | 14 | 15 | def make_parent_dir(filepath): 16 | parent_path = os.path.dirname(filepath) 17 | if not os.path.isdir(parent_path): 18 | try: 19 | os.mkdir(parent_path) 20 | except FileNotFoundError: 21 | make_parent_dir(parent_path) 22 | os.mkdir(parent_path) 23 | print("[INFO] Make new directory: '{}'".format(parent_path)) 24 | 25 | 26 | def save_to_csv(data, path, header=None, index=None): 27 | _assert_suffix_match("csv", path) 28 | make_parent_dir(path) 29 | pd.DataFrame(data).to_csv(path, header=header, index=index) 30 | print("[INFO] Save as csv: '{}'".format(path)) 31 | 32 | 33 | def load_from_csv(path, header=None, index_col=None): 34 | return pd.read_csv(path, header=header, index_col=index_col) 35 | 36 | 37 | def save_to_excel(data, writer, other_kws={}): 38 | if type(writer) == str: # if path is ExcelWriter, skip this validation 39 | _assert_suffix_match("xlsx", writer) 40 | make_parent_dir(writer) 41 | else: 42 | _assert_suffix_match("xlsx", writer.path) 43 | make_parent_dir(writer.path) 44 | if not hasattr(data, "to_excel"): 45 | data = pd.DataFrame(data) 46 | data.to_excel(writer, **other_kws) 47 | print("[INFO] Save as excel: '{}'".format( 48 | writer if type(writer) == str else writer.path 49 | )) 50 | 51 | 52 | def load_from_excel(path, header=[0], index_col=[0]): 53 | return pd.read_excel(path, header=header, index_col=index_col) 54 | 55 | 56 | def save_to_pkl(data, path): 57 | _assert_suffix_match("pkl", path) 58 | make_parent_dir(path) 59 | with open(path, "wb") as f: 60 | pickle.dump(data, f, protocol=-1) 61 | print("[INFO] Save as pkl: '{}'".format(path)) 62 | 63 | 64 | def load_from_pkl(path): 65 | with open(path, "rb") as f: 66 | return pickle.load(f) 67 | 68 | 69 | def save_to_npy(data, path): 70 | _assert_suffix_match("npy", path) 71 | make_parent_dir(path) 72 | np.save(path, data) 73 | print("[INFO] Save as npy: '{}'".format(path)) 74 | 75 | 76 | def load_from_npy(path): 77 | return np.load(path) 78 | 79 | 80 | def save_to_pth(data, path, model=True): 81 | _assert_suffix_match("pth", path) 82 | make_parent_dir(path) 83 | if model: 84 | if hasattr(data, "module"): 85 | data = data.module.state_dict() 86 | else: 87 | data = data.state_dict() 88 | torch.save(data, path) 89 | print("[INFO] Save as pth: '{}'".format(path)) 90 | 91 | 92 | def load_from_pth(path): 93 | return torch.load(path) 94 | 95 | 96 | def save_to_tiff(data, path): 97 | _assert_suffix_match("tiff?", path) 98 | make_parent_dir(path) 99 | tiff.imsave(path, data) 100 | print("[INFO] Save as tiff: '{}'".format(path)) 101 | 102 | 103 | def load_from_tiff(path): 104 | return tiff.imread(path) 105 | 106 | 107 | def savefig_png(path, dpi=150): 108 | _assert_suffix_match("png", path) 109 | make_parent_dir(path) 110 | plt.savefig(path, bbox_inches="tight", dpi=dpi) 111 | print("[INFO] Save figure as png: '{}'".format(path)) 112 | 113 | 114 | def savefig_eps(path): 115 | _assert_suffix_match("eps", path) 116 | make_parent_dir(path) 117 | plt.savefig(path, bbox_inches="tight") 118 | print("[INFO] Save figure as eps: '{}'".format(path)) 119 | 120 | 121 | def saveimg_png(data, path, dpi=150): 122 | _assert_suffix_match("png", path) 123 | make_parent_dir(path) 124 | plt.imsave(fname=path, arr=data, dpi=dpi) 125 | print("[INFO] Save image as png: '{}'".format(path)) 126 | -------------------------------------------------------------------------------- /utils/logger.py: -------------------------------------------------------------------------------- 1 | import time 2 | import logging 3 | from colorama import Fore, Back, Style 4 | 5 | 6 | class ColoredFormatter(logging.Formatter): 7 | def __init__(self): 8 | self._fmt = "{prefix_style}[{levelname:.1} {asctime}]{reset} {message_style}{message}{reset}" 9 | self._style = "{" 10 | super().__init__(fmt=self._fmt, datefmt="%Y-%m-%d %H:%M:%S", style=self._style) 11 | self._pallet = { 12 | "CRITICAL": Fore.BLUE, 13 | "ERROR": Fore.RED, 14 | "WARNING": Fore.YELLOW, 15 | "INFO": Fore.GREEN, 16 | "DEBUG": Fore.MAGENTA, 17 | } 18 | 19 | def format(self, record): 20 | record.prefix_style = self._pallet[record.levelname] 21 | record.message_style = self._pallet[record.levelname]+Style.BRIGHT 22 | record.reset = Style.RESET_ALL 23 | return logging.Formatter.format(self, record) 24 | 25 | 26 | class PrettyLogger(object): 27 | """Customized logger for pretty logging messages""" 28 | 29 | def __init__(self, level=logging.DEBUG): 30 | self._level = level 31 | self.logger = self._init_logger() 32 | 33 | def get_logger(self): 34 | return self.logger 35 | 36 | def _init_logger(self, clear_prev_handlers=True): 37 | logger = logging.getLogger() 38 | logger.setLevel(self._level) 39 | console_handler = logging.StreamHandler() 40 | console_handler.setLevel(self._level) 41 | colored_formatter = ColoredFormatter() 42 | console_handler.setFormatter(colored_formatter) 43 | if clear_prev_handlers: 44 | logger.handlers.clear() 45 | logger.addHandler(console_handler) 46 | return logger 47 | 48 | def _join_words(self, words): 49 | return " ".join(map(str, words)) 50 | 51 | def critical(self, *words): 52 | self.logger.critical(self._join_words(words)) 53 | 54 | def error(self, *words): 55 | self.logger.error(self._join_words(words)) 56 | 57 | def warning(self, *words): 58 | self.logger.warning(self._join_words(words)) 59 | 60 | def info(self, *words): 61 | self.logger.info(self._join_words(words)) 62 | 63 | def debug(self, *words): 64 | self.logger.debug(self._join_words(words)) -------------------------------------------------------------------------------- /utils/timer.py: -------------------------------------------------------------------------------- 1 | from datetime import datetime 2 | 3 | 4 | def record_time(record_list, func, args, time_format="%Y%m%d-%H:%M:%S"): 5 | start_time = datetime.now() 6 | result = func(*args) 7 | end_time = datetime.now() 8 | duration = end_time - start_time 9 | record_list.append([ 10 | start_time.strftime(time_format), 11 | end_time.strftime(time_format), 12 | format_timedelta(duration), 13 | ]) 14 | return result 15 | 16 | def format_timedelta(timedelta): 17 | total_seconds = int(timedelta.total_seconds()) 18 | hours, remainder = divmod(total_seconds, 60*60) 19 | minutes, seconds = divmod(remainder, 60) 20 | return "{}:{}:{}".format(hours, minutes, seconds) 21 | 22 | 23 | --------------------------------------------------------------------------------