Commit Graph

936 Commits

Author SHA1 Message Date
4bd0be410f One more try 2018-03-28 03:27:35 +03:00
42a945fc2e Another fix 2018-03-28 03:20:21 +03:00
196f223888 Compile fix 2018-03-28 03:09:28 +03:00
d9ae3dd681 Added Yolo v3 2018-03-28 02:59:03 +03:00
47c7af1cea Update Readme.md 2018-03-27 01:58:48 +03:00
2302dbed57 Merge pull request #505 from judwhite/patch-2
README.md: add notes to How to compile on Windows
2018-03-27 01:55:33 +03:00
11e6f4b081 Merge pull request #492 from sivagnanamn/master
Minor changes to keep MAC OS happy
2018-03-27 01:48:28 +03:00
bcdf943ca7 reversed order
- oops.
2018-03-25 15:51:48 -05:00
8b7121b34a fix link to issue #500
#nnn alone doesn't work in readme/wiki markdown
2018-03-25 15:50:20 -05:00
c6ecf1e042 README.md: add notes to How to compile on Windows 2018-03-25 15:41:48 -05:00
8f1f5cbf83 3.5x times accelerated data augmentation using OpenCV for training Yolo on DGX and multi-GPU 2018-03-24 23:35:05 +03:00
013031ce3f Minor changes to keep MAC OSX happy 2018-03-21 13:54:44 +09:00
101de2b07a More stable web-cam 2018-03-21 00:40:01 +03:00
fe4f01b1d8 Fixed uses of WebCam for OpenCV 3.x 2018-03-20 22:10:46 +03:00
3e5abe0680 Update Readme.md 2018-03-20 03:14:20 +03:00
ced198e939 Fixed gpu_id for DLL/SO 2018-03-20 02:17:26 +03:00
537d135feb Improve training performance - batch-norm using cuDNN. 2018-03-20 02:16:51 +03:00
2f52cfeb07 Fixed compile error 2018-03-17 01:12:00 +03:00
e947bf29af C wrappers 2018-03-17 01:03:56 +03:00
164ba34279 Fixed bug, that save weights-file for each iteration 2018-03-16 14:58:21 +03:00
371f21171a shortcut_layer resize for random=1 2018-03-16 01:49:01 +03:00
df076653e0 Minor fix 2018-03-13 19:04:39 +03:00
15c89e7a71 Added include for sprintf 2018-03-12 14:11:53 +03:00
880cf187d8 Fixed multi-GPU training for Tensor Cores 2018-03-09 19:44:46 +03:00
a6c51e3b75 Fixes for small objects. And fixes of densenet201_yolo.cfg, resnet50_yolo.cfg, resnet152_yolo.cfg 2018-03-09 01:42:58 +03:00
100d6f7801 Remove 1 additional conv 3x3 kernel layer from resnet152_yolo.cfg and resnet50_yolo.cfg 2018-03-06 02:38:54 +03:00
ae1768e583 Removed random=1 from resnet152_yolo.cfg. Until resize_network() isn't supported for [shortcut] layer 2018-03-05 23:26:09 +03:00
e90afa6ec1 resnet152_yolo.cfg uses learning_rate, steps, scales, decay, momentum as in the RetinaNet 2018-03-05 23:07:33 +03:00
0c07837397 Added resnet152_yolo.cfg with focal_loss=1 and stopbackward=1 in the last shortcut layer. 2018-03-05 20:31:43 +03:00
5e448c00f3 Added Focal loss. Use focal_loss=1 in the [region] layer in your cfg-file.
alpha = 0.5, gamma = 2
2018-03-05 19:59:29 +03:00
f0bc634a74 Fixed saving video file 2018-03-05 13:25:27 +03:00
09fdb67d66 Update Readme.md 2018-03-05 01:43:56 +03:00
c123f320fc Show avg-loss chart during Training, if compiled with OpenCV. Use -dont_show to disable. 2018-03-04 23:41:23 +03:00
48586c8d4d Compile fix 2018-03-04 18:15:56 +03:00
74f98da211 Show IoU, save anchors to file.Show anchors in the window if used -show 2018-03-04 17:59:23 +03:00
f39160f6e8 Added: calc_anchors 2018-03-04 03:09:40 +03:00
fe44d3d0f2 Fixed reorg_layer. Old reorg can be used as [reorg_old] in the cfg-file. 2018-03-03 18:09:46 +03:00
e16c901163 minor fix 2018-03-02 00:16:26 +03:00
be9ccb7d7f Added F1 score to accuracy statistic 2018-03-02 00:09:52 +03:00
1e84a2804e Update Readme.md 2018-03-01 18:38:58 +03:00
671735fb00 Update Readme.md 2018-03-01 18:36:14 +03:00
7caf345d36 One more fix 2018-03-01 16:42:30 +03:00
18d5670d64 Compile fix 2018-03-01 16:33:40 +03:00
5cd147031b Added param -dont_show to detect objects on servers without display, like Amazon EC2 2018-03-01 16:05:39 +03:00
35a66b73a2 Changed class to class_id 2018-03-01 15:14:58 +03:00
17b22d7ce7 Update Readme.md 2018-03-01 03:10:32 +03:00
490d02505b Fixed VideoWriter 2018-02-26 21:48:09 +03:00
e96a454ca1 Update Readme.md 2018-02-26 01:13:55 +03:00
48d713fd49 Fine-tuning densenet201_yolo.cfg and resnet50_yolo.cfg 2018-02-26 01:09:14 +03:00
cad4d1618f Added support for Tensor Cores CC >= 7.0 (V100). For FP16/32 (mixed precision) define CUDNN_HALF should be used. 2018-02-25 16:29:44 +03:00