Commit Graph

Select branches

Hide Pull Requests

master

#6

Mono Color

• 031a1f9ef0 Removed experimental AUTOTUNE Jim Martens 2019-04-15 12:14:06 +02:00
• d6e3cba225 Piped verbose option through to train Jim Martens 2019-04-15 12:13:37 +02:00
• 1243cc0a4b Removed duplicate loss log Jim Martens 2019-04-15 10:24:16 +02:00
• 3afb74587b Updated image paths to work with cvpc7 installation Jim Martens 2019-04-15 10:21:36 +02:00
• c8e49bd0da Removed first deconv layer to achieve symmetry in convolutions Jim Martens 2019-04-14 09:36:10 +02:00
• 1764e10da4 Removed unnecessary layers and fixed the names of the remaining ones Jim Martens 2019-04-12 14:26:43 +02:00
• 32aecaea36 Deactivated pooling and upsampling layers Jim Martens 2019-04-12 14:21:47 +02:00
• 1ea9ff98d4 Increased zsize to 64 Jim Martens 2019-04-12 14:17:17 +02:00
• 24e8d0e009 Deactivated dropout Jim Martens 2019-04-12 14:17:05 +02:00
• d749678d20 Removed unnecessary layers Jim Martens 2019-04-12 14:16:48 +02:00
• 568adeab08 Implemented auto-encoder model from OliverEdholm Jim Martens 2019-04-12 13:45:27 +02:00
• 7ea63a9a49 Reverted to default AdamOptimizer beta values Jim Martens 2019-04-12 13:44:05 +02:00
• 25e2110a5b Changed size of images to 256, 256 Jim Martens 2019-04-12 13:43:29 +02:00
• 37b0b702a3 Changed batch size to 16 Jim Martens 2019-04-12 13:42:48 +02:00
• 1733fc8c99 Added data directory to git ignore Jim Martens 2019-04-12 13:39:20 +02:00
• 30d2e156fe Reduced zsize and increases learning rate Jim Martens 2019-04-11 15:20:48 +02:00
• e7be92550d Resized images to 64x64 to alleviate memory problems Jim Martens 2019-04-11 15:20:28 +02:00
• 41fe940d5b Modified model as per instructions from Keras blog Jim Martens 2019-04-11 15:19:50 +02:00
• 0d0ea882dd Made filter sizes in encoder/decoder models dependent on variable Jim Martens 2019-04-11 13:31:06 +02:00
• 28ac1a2027 Fixed mapping from image to bounding box Jim Martens 2019-04-10 17:42:59 +02:00
• a0f659cea2 Cleaned up data set to prevent awkward cases Jim Martens 2019-04-10 17:00:32 +02:00
• b173746784 Ensured summaries are written Jim Martens 2019-04-10 15:19:44 +02:00
• a7f9d586f4 Fixed loop through data set Jim Martens 2019-04-10 15:18:39 +02:00
• da6e348edc Fixed data loading Jim Martens 2019-04-10 15:17:24 +02:00
• f81ecbac0d Fixed slicing Jim Martens 2019-04-10 15:16:13 +02:00
• f30edeb3f2 Added command line interface Jim Martens 2019-04-09 14:09:22 +02:00
• ab24c96515 Fixed required packages Jim Martens 2019-04-09 14:07:42 +02:00
• 523a9c8e71 Removed hard-coded reliance on specific batch size Jim Martens 2019-04-09 10:33:14 +02:00
• be501d606f Modified data module to cut everything but the bounding box Jim Martens 2019-04-09 10:18:50 +02:00
• 57da44c6d8 Modified data loader to supply the full trainval35k data Jim Martens 2019-04-04 19:04:19 +02:00
• a8ef3d000f Added argument to specify the inlying class Jim Martens 2019-04-04 17:35:11 +02:00
• d499c55ab4 Added skeleton function for loading SceneNet RGB-D data Jim Martens 2019-04-04 17:30:52 +02:00
• d267512312 Added info about default value to docstring Jim Martens 2019-04-04 17:30:32 +02:00
• 05fa489905 Removed sample generation Jim Martens 2019-04-04 17:28:27 +02:00
• 17537b944b Removed left-over reference to early stopping Jim Martens 2019-04-04 17:26:56 +02:00
• 9a1f91892a Fixes typo Jim Martens 2019-04-04 17:26:06 +02:00
• 1f25122749 Removed reference to early stopping from docstring Jim Martens 2019-04-04 17:25:34 +02:00
• c83a65571b Corrected docstrings Jim Martens 2019-04-04 17:24:05 +02:00
• 53955cbc46 Removed imports of tensorflow shortcuts Jim Martens 2019-04-04 17:22:07 +02:00
• 08f26fa9e0 Extracted preparation of MNIST data into separate file Jim Martens 2019-04-04 17:21:29 +02:00
• 403577da0a Extracted adversarial auto-encoder functions into separate file Jim Martens 2019-04-04 17:14:11 +02:00
• 655b3a63d4 Converted to f strings Jim Martens 2019-04-04 17:00:09 +02:00
• df733e69c4 Optimised imports Jim Martens 2019-04-04 16:57:03 +02:00
• 84bd2c2e7b Added data loader for COCO data set Jim Martens 2019-04-04 16:56:50 +02:00
• f6699b8444 Added pycocotools as dependency Jim Martens 2019-04-04 14:23:30 +02:00
• 05ea13d252 Moved COCO wrapper into package Jim Martens 2019-04-01 12:54:58 +02:00
• 855d8764d0 Removed obsolete log outputs Jim Martens 2019-03-30 12:12:27 +01:00
• 8ef7d5f999 Removed obsolete parameters Jim Martens 2019-03-25 14:33:41 +01:00
• 6835c284af Fixed coding style Jim Martens 2019-03-25 14:32:08 +01:00
• 9a08ea3bb9 Added simple auto-encoder training Jim Martens 2019-03-25 14:31:26 +01:00
• 0eab8cd123 Added mapping from COCO to WNID Jim Martens 2019-03-25 13:43:29 +01:00
• db0d275bdd Save predictions batch-wise to prevent memory issues Jim Martens 2019-03-21 17:49:34 +01:00
• ae109fd131 Switched output saving to numpy format Jim Martens 2019-03-21 17:38:22 +01:00
• 69cc07ed12 Corrected type hinting Jim Martens 2019-03-21 17:31:42 +01:00
• 976563526a Added train and predict functions to module documentation Jim Martens 2019-03-21 17:14:44 +01:00
• 7ba3333e28 Updated reference to ssd_keras repository Jim Martens 2019-03-21 17:12:35 +01:00
• 9eb1283b8b Added prediction functionality to ssd module Jim Martens 2019-03-21 17:12:17 +01:00
• f40d04e34c Updated reference to new ssd_keras version Jim Martens 2019-03-20 16:53:04 +01:00
• 5a7bd6ea9d Fixed image size constant Jim Martens 2019-03-20 16:52:47 +01:00
• b7825c0095 Implemented training function for SSD Jim Martens 2019-03-20 16:46:14 +01:00
• 167899b2f8 Added iou threshold and top_k parameters to function calls Jim Martens 2019-03-20 14:55:32 +01:00
• 9a7a906acf Added ability to load trained weights Jim Martens 2019-03-20 14:54:28 +01:00
• 6ac3b5fdcf Added SSD and DropoutSSD classes Jim Martens 2019-03-20 14:34:46 +01:00
• a9d73a1f47 Added empty ssd module Jim Martens 2019-03-20 13:17:56 +01:00
• c090e2c100 Added subpackage info about SSD Jim Martens 2019-03-20 13:15:58 +01:00
• b9519a516d Added dropout SSD and type hinting for used models Jim Martens 2019-03-20 13:13:55 +01:00
• 262a36a68b Added ssd_keras sub-module Jim Martens 2019-03-19 17:21:09 +01:00
• be4ee74537 Updated README with badges, links and some text Jim Martens 2019-02-08 23:40:45 +01:00
• 26904614cb Improved formatting of LICENSE file Jim Martens 2019-02-08 22:59:17 +01:00
• dddd63aabc Added boilerplate README (needs to be changed) Jim Martens 2019-02-08 22:57:41 +01:00
• c7a78c35a4 Added package documentation to masterthesis Jim Martens 2019-02-08 22:56:58 +01:00
• 416cb4c7b5 Reformatted code to ensure all parameters are on their own line Jim Martens 2019-02-08 22:42:49 +01:00
• 9909b5fdfc Improved docstrings in model module Jim Martens 2019-02-08 22:42:23 +01:00
• 1e422d3fa7 Improved imports according to Google standards Jim Martens 2019-02-08 22:37:29 +01:00
• 422e01a917 Improved docstring in aae package Jim Martens 2019-02-08 22:23:24 +01:00
• 85f3100b5d Updated docstring for model module Jim Martens 2019-02-08 21:57:57 +01:00
• 81f245be99 Updated docstring for util module Jim Martens 2019-02-08 21:22:00 +01:00
• 9bf7229b1e Added TODOs to train module Jim Martens 2019-02-08 21:14:45 +01:00
• 90fa27d9ef Replaced all binary cross entropy losses with log_loss Jim Martens 2019-02-08 21:01:38 +01:00
• a51b535728 Improved naming for histogram summaries and added variable summaries Jim Martens 2019-02-08 20:47:34 +01:00
• 0cbab44f01 Fixed summaries for histograms Jim Martens 2019-02-08 20:25:56 +01:00
• a319cf988e Added summaries for grads and learning rate Jim Martens 2019-02-08 20:13:21 +01:00
• e85b850c63 Fixed frequency of summary savings Jim Martens 2019-02-08 19:41:08 +01:00
• a8f8015ad8 Improved summary logs Jim Martens 2019-02-08 19:27:33 +01:00
• a2aa794d47 Fixed sample image summary Jim Martens 2019-02-08 18:58:19 +01:00
• 3b9742a1b4 Improved summary logging Jim Martens 2019-02-08 18:40:41 +01:00
• 75d7c769c7 Made early stopping conditional and turned it off Jim Martens 2019-02-08 18:22:25 +01:00
• ea1b4943cf Fixed function call Jim Martens 2019-02-08 18:12:33 +01:00
• d9a35893af Added epoch counter to restored variables Jim Martens 2019-02-08 18:12:02 +01:00
• 1d4f56e40d Improved docstrings Jim Martens 2019-02-08 17:12:41 +01:00
• 028513b404 Renamed internal functions to make them protected Jim Martens 2019-02-08 16:38:58 +01:00
• 483cb4eb3e Removed obsolete function Jim Martens 2019-02-08 16:35:00 +01:00
• 2d1fee8048 Moved training of epoch into separate function Jim Martens 2019-02-08 16:34:23 +01:00
• 6f36aa7faf Added more parameters to checkpoint Jim Martens 2019-02-08 15:45:27 +01:00
• 658973d90d Added weights prefix to parameters Jim Martens 2019-02-08 15:34:49 +01:00
• 88ffd4f879 Moved dataset preparation into own function Jim Martens 2019-02-08 15:32:01 +01:00
• 8e0a063759 Removed abspath usage in specifying checkpoint dir Jim Martens 2019-02-08 15:03:47 +01:00
• 279e848434 Fixed specification of checkpoint dir Jim Martens 2019-02-08 14:59:50 +01:00
• 720ff63c7d Updated saving of checkpoints Jim Martens 2019-02-08 14:53:45 +01:00
• 3c5f332880 Added saving of weights after each epoch Jim Martens 2019-02-08 14:22:27 +01:00