github/DeepFaceLab
1
0
Fork 0
mirror of https://github.com/iperov/DeepFaceLab.git synced 2025-07-07 13:32:09 -07:00
Code Issues Projects Releases Packages Wiki Activity Actions
1242 commits 3 branches 2 tags 866 MiB
Commit graph

148 commits

Author SHA1 Message Date
iperov
4581d48ff0 _ 2021-08-10 20:54:54 +04:00
iperov
55b947eab5 XSeg: added pretrain option. 2021-07-30 17:24:21 +04:00
iperov
f044c99ddc remove ConvolutionAwareInitializer from layers, because lr-dropout does the same thing. 2021-07-15 00:50:11 +04:00
iperov
2edac3df8c fix for clipgrad 2021-07-15 00:24:38 +04:00
iperov
35877dbfd7 AMP, SAEHD: added option use_fp16: Increases training/inference speed, reduces model size. Model may crash. Enable it after 1-5k iters. 
AMP: removed lr_dropout, now it is enabled by default;
 2021-07-14 23:03:54 +04:00
iperov
ee1bc83a14 AMP, SAEHD: reverted GAN to december version. 2021-07-07 21:45:43 +04:00
iperov
62c6fffdde fix 2021-06-10 14:42:58 +04:00
iperov
5dc027a8b0 _ 2021-06-09 19:17:18 +04:00
iperov
e52b53f87c AMP fix 2021-05-30 09:24:23 +04:00
iperov
e6e2ee7466 pixel_norm op 2021-05-25 14:26:48 +04:00
iperov
66bb72f164 XSeg model has been changed to work better with large amount of various faces, thus you should retrain existing xseg model. 
Windows build: Added Generic XSeg model pretrained on various faces. It is most suitable for src faceset because it contains clean faces, but also can be applied on dst footage without complex face obstructions.
 2021-05-12 13:28:48 +04:00
iperov
081d8faa45 XSeg sample generator: additional sample augmentation for training 2021-04-30 16:49:32 +04:00
iperov
e53d1b1820 XSeg trainer: added random relighting sample augmentation to improve generalization 2021-04-26 10:51:06 +04:00
iperov
23130cd56a fix for colab 2021-04-22 21:12:42 +04:00
iperov
fdb143ff47 added AMD/Intel cards support via DirectX12 ( DirectML backend ) 2021-04-22 18:19:15 +04:00
iperov
dcf146cc16 imagelib : apply_random_sharpen 2021-04-22 18:16:42 +04:00
iperov
65432d0c3d imagelib: random crop func (unused) 2021-04-15 21:43:18 +04:00
iperov
457a39c093 mathlib update 2021-04-10 09:58:33 +04:00
iperov
243f73fafc core.imagelib: update 2021-04-10 09:58:14 +04:00
iperov
f387179cba Added faceset resize tool via 
4.2) data_src util faceset resize.bat
5.2) data_dst util faceset resize.bat

Resize faceset to match model resolution to reduce CPU load during training.

Don’t forget to keep original faceset.
 2021-03-23 16:01:24 +04:00
iperov
11add4cd4f update GAN model 2021-01-03 19:25:39 +04:00
iperov
54fc3162ed new gan 2021-01-03 01:08:20 +04:00
iperov
140f16f772 Merge branch 'master' into new_gan 2021-01-01 18:04:10 +04:00
iperov
4f2efd7985 fix support for v1/v2 2021-01-01 17:59:57 +04:00
iperov
8ff34be5e4 leras.nn : support for tf ver 1 2021-01-01 17:37:12 +04:00
iperov
ad5733c5bb upd PatchDiscriminator 2020-12-31 17:38:41 +04:00
iperov
ae9e16b4a5 SAEHD: new GAN 2020-12-30 14:33:14 +04:00
iperov
241d1a9c35 leras ops: pixelnorm, total_variation_mse 2020-12-30 14:32:07 +04:00
iperov
299d91c81e remove unused 2020-12-30 14:31:23 +04:00
iperov
cc950f12fb set CUDA_​CACHE_​MAXSIZE to 2gb before first cuda load 2020-12-28 18:37:29 +04:00
iperov
b7bed0ef5e optimize MPSharedList.py 2020-12-20 23:04:00 +04:00
iperov
977d8a2d77 The load time of training data has been reduced significantly. 2020-12-20 19:17:24 +04:00
iperov
e7d36b4287 leras: added AdaBelief optimizer 2020-12-11 15:47:47 +04:00
iperov
35945b257c fix dimensions error 2020-12-11 14:34:37 +04:00
iperov
b9c9e7cffd fix depth_to_space for tf2.4.0. Removing compute_output_shape in leras, because it uses CPU device, which does not support all ops. 2020-12-11 11:28:33 +04:00
Colombo
bb432b21f9 . 2020-11-25 12:19:43 +04:00
Colombo
c516454566 fix rmsprop 2020-11-18 13:59:14 +04:00
Colombo
874a7eba18 1 2020-11-18 13:41:19 +04:00
Colombo
1adad3ece6 fix RMSprop.py 2020-11-18 13:38:27 +04:00
Colombo
0eb7e06ac1 upgrade to tf 2.4.0rc1 2020-11-13 17:00:07 +04:00
Colombo
dd21880ecd fix 2020-07-16 23:39:56 +04:00
Colombo
e8b04053e4 SAEHD: 
Changed help for “Learning rate dropout” option:

When the face is trained enough, you can enable this option to get extra sharpness and reduce subpixel shake for less amount of iterations.

Enabled it before “disable random warp” and before GAN. n disabled. y enabled

cpu enabled on CPU. This allows not to use extra VRAM, sacrificing 20% time of iteration.

 Changed help for GAN option:

Train the network in Generative Adversarial manner.

Forces the neural network to learn small details of the face.

Enable it only when the face is trained enough and don't disable.

Typical value is 0.1

improved GAN. Now it produces better skin detail, less patterned aggressive artifacts, works faster.
https://i.imgur.com/Nbh3mw1.png
 2020-07-16 22:19:17 +04:00
Colombo
aacd29269a fix 2020-07-16 22:11:23 +04:00
Colombo
5a40f537cc . 2020-07-04 07:58:53 +04:00
Colombo
9a540e644c upd leras ops 2020-07-03 19:32:14 +04:00
Colombo
b0ad36de94 upd cv2ex 2020-07-03 19:30:52 +04:00
Colombo
3283dce96a interact: change to waitKeyEx 2020-07-03 18:43:18 +04:00
Colombo
770da74a9b fix cv2.resize interpolation 2020-07-03 18:40:35 +04:00
Colombo
4ce4997d1a fix 2020-07-03 15:19:12 +04:00
Colombo
e6aa996814 leras: add ability to save sub layers in a dict 2020-07-01 22:17:35 +04:00