fanwuyang
/
neural-enhance-super-resolution
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Commit Graph

  • 71b0dba9f3 Initial Commit main Fan-Wu Geoffrey Yang 2023-01-13 13:02:47 +1300
  • 2fd67de7ed Merge pull request #108 from concutere/fix-load-model-absolute-path Alex J. Champandard 2016-11-30 16:15:55 +0100
  • 5248437d06 fix load_model to use absolute path concutere 2016-11-30 06:28:18 -0800
  • d1ab2b1fa3 Merge pull request #94 from m-tr/fix/dockerfile-wget Alex J. Champandard 2016-11-19 20:27:56 +0100
  • 14e1734a56 Dockerfiles wget fix m-tr 2016-11-19 19:42:47 +0100
  • b3a9ba58c7 Merge pull request #93 from AlexeyKruglov/fix-hist Alex J. Champandard 2016-11-19 12:29:08 +0100
  • 8e60e74978 Fix histogram matching with scipy 0.17.0 Alexey Kruglov 2016-11-18 21:24:17 +0300
  • 4ee13f5ac5 Add ne4x pretrained weights to docker image (#85) fchouteau 2016-11-16 16:52:17 +0100
  • 559e66bbb2 Fix exception when training with JPEG compression. Alex J. Champandard 2016-11-12 15:47:00 +0100
  • 9d2aa3c436 Fix histogram colors overflowing in certain images. Closes #73. Alex J. Champandard 2016-11-12 15:43:23 +0100
  • 904071d993 Fix memory error in Docker by changing default tile size. Closes #74. Alex J. Champandard 2016-11-12 15:41:30 +0100
  • 9fae68b69e Moved demo website to its own URL on nucl.ai rather than IP. Alex J. Champandard 2016-11-11 15:01:33 +0100
  • 4000bd365c Fix filenames used on stdout to relative paths. Alex J. Champandard 2016-11-11 10:29:06 +0100
  • 881e909dc5 Fix default parameters to zoom=2 so it works without arguments. Closes #71. Alex J. Champandard 2016-11-11 10:06:59 +0100
  • a22d7ced3c Rename folder with training script so they are not used for enhancing by users accidentally. Alex J. Champandard 2016-11-11 10:00:23 +0100
  • fabe04f1d0 Fix bug in handling of --train-jpeg parameter. Alex J. Champandard 2016-11-11 01:33:06 +0100
  • 7c8cc5609f Add training scripts, they take a while to run! Alex J. Champandard 2016-11-11 00:33:54 +0100
  • 9d8ab0ea8e Merge pull request #69 from alexjc/v0.3 Alex J. Champandard 2016-11-10 17:29:18 +0100
  • 80d7b57bec Prepare build scripts for merge and update README. Alex J. Champandard 2016-11-10 17:20:20 +0100
  • b706599cb1 Fix alias for bash and zsh, Closes #54. Add standalone alias-free docker example, Closes #46. Alex J. Champandard 2016-11-10 17:06:52 +0100
  • c03a0b7fa3 Fix for trained models loaded from script directory. Closes #67. Alex J. Champandard 2016-11-10 16:55:56 +0100
  • a69d2ae060 Add histogram color matching, enabled with `--rendering-histogram`. Alex J. Champandard 2016-11-10 16:45:30 +0100
  • 965bd55090 Add new model selection (by image type and 'enhance' operation) and parameters and documentation in README. Alex J. Champandard 2016-11-10 13:16:40 +0100
  • e6120878c6 Minor rework of the training parameters. Alex J. Champandard 2016-11-09 17:04:35 +0100
  • 0d4621df0a Simplify architecture outside of convolution blocks for speed, improve training code. Alex J. Champandard 2016-11-08 17:41:28 +0100
  • c8e805667f Fix for padding in convolution. Alex J. Champandard 2016-11-08 17:15:54 +0100
  • 2a29fc954b Buffer chunks of images proportionally to size. Alex J. Champandard 2016-11-08 17:15:41 +0100
  • da354f3e55 Remove reflection layer, important but not the bottleneck right now and quite slower. Alex J. Champandard 2016-11-04 14:29:15 +0100
  • 9bba9850d5 Improve pre-processing steps to include more randomness: scales, blur... Alex J. Champandard 2016-11-04 14:28:37 +0100
  • 149ae4e566 Update training parameters. Alex J. Champandard 2016-11-04 00:34:52 +0100
  • de047d6a74 Fix for images that are not multiples of 2 or 4, depending on the rescaling. Alex J. Champandard 2016-11-04 00:32:08 +0100
  • 29bfe58d73 Merged documentation changes. Alex J. Champandard 2016-11-04 00:03:21 +0100
  • 9a228c1d78 Merge pull request #44 from wieczorek1990/master Alex J. Champandard 2016-11-03 23:58:24 +0100
  • b24fa7490d Merge pull request #47 from alexjc/v0.2 Alex J. Champandard 2016-11-03 22:33:15 +0100
  • 03914db364 Fix formatting, minor tweaks to Docker build files for release. Alex J. Champandard 2016-11-03 22:27:28 +0100
  • 448e7b93dc Fix progress output when tiled rendering, changed default zoom level. Alex J. Champandard 2016-11-03 22:01:08 +0100
  • cabaaeeefe Add training scripts for networks currently being trained, for release v0.2. Alex J. Champandard 2016-11-03 20:12:58 +0100
  • 1ad40b6d71 Merge branch 'master' into v0.2 Alex J. Champandard 2016-11-03 19:27:10 +0100
  • ac49676415 Add tiled rendering with padding, no feather-blending but looks good enough. Alex J. Champandard 2016-11-03 19:26:55 +0100
  • ed20dcefd5 Fixed typo Łukasz Wieczorek 2016-11-03 12:05:33 +0100
  • 095fe42dc3 Add tiled rendering, currently with no padding for each tile. Alex J. Champandard 2016-11-03 10:35:43 +0100
  • d18c08f1b5 Integrated reflection padding instead of zero padding for extra quality during training and inference. Alex J. Champandard 2016-11-02 21:39:34 +0100
  • 90c0b7ea43 Fix padding code, more reliable for specific upscale/downscale combinations. Alex J. Champandard 2016-11-02 13:03:00 +0100
  • 3b2a6b9d8d Add extra padding on input to avoid zero-padding. Experiment with training values from ENet (segmentation). Alex J. Champandard 2016-11-02 12:08:07 +0100
  • 7924cc4a85 Improve display and filenames for saving output. Alex J. Champandard 2016-11-02 10:16:31 +0100
  • 93e5a41d9a Fix and optimize pre-processing of images. Alex J. Champandard 2016-11-01 20:40:10 +0100
  • 11ba505252 Fix for gradient clipping code. Alex J. Champandard 2016-11-01 16:04:33 +0100
  • cf65207a2e Use full range of tanh output rather than [-0.5, +0.5], avoids clipping. Alex J. Champandard 2016-11-01 16:02:18 +0100
  • 0c31e53731 Fix suggested alias for relative paths. Closes #37 #28. Alex J. Champandard 2016-11-01 15:11:55 +0100
  • 34f8e629c2 Improve the alias used to invoke docker, so it's more robust to directory locations and input paths. Alex J. Champandard 2016-11-01 09:54:58 +0100
  • c610623b11 Add gradient clipping, helpful for preventing problems with extreme parameters/architectures. Alex J. Champandard 2016-10-31 21:38:23 +0100
  • 02d2fca6c5 Corrected value for adversarial loss. Don't refactor math the day after stopping coffee. Alex J. Champandard 2016-10-31 20:55:11 +0100
  • f2494f8078 Add new downscale layers, separate from upscale steps. Renamed --scales to --zoom for inference. Alex J. Champandard 2016-10-31 13:49:17 +0100
  • 0c9937a317 Merge pull request #22 from msfeldstein/master Alex J. Champandard 2016-10-31 13:29:31 +0100
  • 064f9dd589 Add three image pre-processing options, improve loading code. Alex J. Champandard 2016-10-31 06:21:52 +0100
  • fef84c5b44 Remove cnmem theano flag since it doesn't work if you're sharing GPU with display. Michael Feldstein 2016-10-30 19:24:50 -0700
  • 5ef872b876 Add warning for files that may be too large for 4x. Alex J. Champandard 2016-10-30 20:45:10 +0100
  • a5ad2c25e6 Merge pull request #18 from alexjc/training Alex J. Champandard 2016-10-30 20:29:13 +0100
  • 17fcad8d28 Refactor of changes related to training. Alex J. Champandard 2016-10-30 20:26:03 +0100
  • 2b67daedb6 Merge pull request #12 from dribnet/generic_seeds Alex J. Champandard 2016-10-30 17:19:33 +0100
  • cad5eff572 Merge pull request #11 from dribnet/save_every_epoch Alex J. Champandard 2016-10-30 17:18:33 +0100
  • a9b0cd9887 Merge pull request #9 from zuphilip/patch-1 Alex J. Champandard 2016-10-30 17:07:15 +0100
  • fcc5e87858 Merge pull request #4 from dribnet/valid_dir Alex J. Champandard 2016-10-30 17:06:33 +0100
  • 1478977f18 Merge pull request #10 from OndraM/fix-duplicate-param Alex J. Champandard 2016-10-30 16:53:48 +0100
  • 8f5167d235 Fix enhancer.process to pass img, seed Tom White 2016-10-31 02:13:02 +1300
  • 37cb208374 Move generation of seeds out of training network Tom White 2016-10-31 01:18:46 +1300
  • b05ee6ad08 Added --save-every-epoch option Tom White 2016-10-31 00:12:28 +1300
  • c5053806bd Add valid dir when necessary Tom White 2016-10-29 23:55:11 +1300
  • eb25e737cf Fix duplicate param definition Ondřej Machulda 2016-10-30 11:48:53 +0100
  • f83e69e96a Fix some typos in README Philipp Zumstein 2016-10-30 11:11:54 +0100
  • f68f04fb1c Improve instructions to train custom models so a new file is output and existing one is not loaded. Use --model parameter! Alex J. Champandard 2016-10-30 10:50:57 +0100
  • 203917d122 Switch default to small model to reduce memory usage. Alex J. Champandard 2016-10-30 10:49:23 +0100
  • 2b5fc8f51d Add docker instructions, fix for slow compute in CPU image. Alex J. Champandard 2016-10-30 02:04:28 +0200
  • 74cb95609e Fix for docker build using latest Miniconda and Python 3.5 explicitly. Alex J. Champandard 2016-10-29 23:57:08 +0200
  • 99c767b7e2 Add docker configuration files for CPU and GPU. Alex J. Champandard 2016-10-29 22:29:30 +0200
  • bf22450b8d Update documentation for new --train usage, minor improvements. Alex J. Champandard 2016-10-29 14:15:24 +0200
  • 4c55c48f62 Add argument for specifying training images, cleaned up file handling. Alex J. Champandard 2016-10-29 12:39:07 +0200
  • 1c38f2ca31 New meme-friendly image and link to demo. Alex J. Champandard 2016-10-29 01:35:44 +0200
  • b1c054ce9f Improve the README for applying and training models. Alex J. Champandard 2016-10-29 00:39:32 +0200
  • f868514be3 Improve code for simply applying super-resolution. Alex J. Champandard 2016-10-29 00:18:44 +0200
  • 30534c6dd1 Add old station example, fix bank example. Alex J. Champandard 2016-10-28 22:55:10 +0200
  • 801a4707f4 Improve README and the visual examples. Alex J. Champandard 2016-10-28 16:29:37 +0200
  • 39e12ea205 Add example GIF to the README, tweak text. Alex J. Champandard 2016-10-28 14:38:28 +0200
  • c456221cb5 Experiment with recursive super-resolution and weight reuse, mixed results. Alex J. Champandard 2016-10-28 04:28:30 +0200
  • 87304c93a6 Use traditional learning rate decaying rather than fast-restarts, works better when training continuously adapting GAN. Alex J. Champandard 2016-10-28 04:27:17 +0200
  • 3809e9b02a Add loading of images into a buffer, using multiple fragments per JPG loaded. Works well with larger datasets like OpenImages, fully GPU bound. Alex J. Champandard 2016-10-28 04:25:10 +0200
  • 619fad7f3c Add loading parameters from saved models. Clean up learning-rate code. Alex J. Champandard 2016-10-11 17:24:10 +0200
  • bb3b24b04c Add README and requirements file. Alex J. Champandard 2016-10-11 17:16:06 +0200
  • f978f058dd Add descriptions for command-line arguments. Alex J. Champandard 2016-10-11 11:12:49 +0200
  • d8c9292b80 Tweak code for generalized super-resolution. Alex J. Champandard 2016-10-11 08:59:42 +0200
  • 9558a11397 Add command-line for super-resolving images without training. Alex J. Champandard 2016-10-07 21:05:38 +0200
  • 03a6813e95 Enable adversarial loss during training. Alex J. Champandard 2016-10-05 14:52:28 +0200
  • 6542f435b4 Stabilize the training, tuned the architecture. Alex J. Champandard 2016-10-05 00:27:26 +0200
  • 4580e5531a Add .gitignore and removed symlink. Alex J. Champandard 2016-10-04 13:19:53 +0200
  • b52530cf20 Add support for PRELU (default), optional load/save. Alex J. Champandard 2016-10-04 11:25:03 +0200
  • 4d82c7f157 Improve console output. Alex J. Champandard 2016-10-04 02:16:34 +0200
  • 8770113f9a Add support for loading/saving models. Fixed regressions. Alex J. Champandard 2016-10-04 01:40:30 +0200
  • 58e1bed4a6 Improve display and code structure. Alex J. Champandard 2016-10-03 22:17:37 +0200
  • 749b467f94 Switch to sub-pixel deconvolution layer. Alex J. Champandard 2016-10-02 21:25:02 +0200
  • 9f13695050 Add discriminator network too. Alex J. Champandard 2016-10-02 14:01:51 +0200