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
568 commits 3 branches 2 tags 866 MiB
Commit graph

190 commits

Author SHA1 Message Date
Colombo
201b762541 SAE: fix 2019-09-14 16:39:47 +04:00
Colombo
b6b92bded0 converter: now writes a filename of current frame config, 
SAE: removed multiscale decoder, because it's not effective
 2019-09-13 08:59:00 +04:00
Colombo
7ed38a8097 Converter: 
Session is now saved to the model folder.

blur and erode ranges are increased to -400+400

hist-match-bw is now replaced with seamless2 mode.

Added 'ebs' color transfer mode (works only on Windows).

FANSEG model (used in FAN-x mask modes) is retrained with new model configuration
and now produces better precision and less jitter
 2019-09-07 13:57:42 +04:00
iperov
6c66495bcf default avatar type option is now HEAD. Head produces stable result. 2019-08-25 08:16:54 +04:00
iperov
00dce38187 fix 2019-08-25 07:43:22 +04:00
iperov
e3505d9b8c fix AVATARModel 2019-08-24 20:33:29 +04:00
iperov
5968ac21f6 fix AVATARModel 2019-08-24 19:55:38 +04:00
iperov
407ce3b1ca Added interactive converter. 
With interactive converter you can change any parameter of any frame and see the result in real time.

Converter: added motion_blur_power param.
Motion blur is applied by precomputed motion vectors.
So the moving face will look more realistic.

RecycleGAN model is removed.

Added experimental AVATAR model. Minimum required VRAM is 6GB (NVIDIA), 12GB (AMD)
Usage:
1) place data_src.mp4 10-20min square resolution video of news reporter sitting at the table with static background,
   other faces should not appear in frames.
2) process "extract images from video data_src.bat" with FULL fps
3) place data_dst.mp4 video of face who will control the src face
4) process "extract images from video data_dst FULL FPS.bat"
5) process "data_src mark faces S3FD best GPU.bat"
6) process "data_dst extract unaligned faces S3FD best GPU.bat"
7) train AVATAR.bat stage 1, tune batch size to maximum for your card (32 for 6GB), train to 50k+ iters.
8) train AVATAR.bat stage 2, tune batch size to maximum for your card (4 for 6GB), train to decent sharpness.
9) convert AVATAR.bat
10) converted to mp4.bat

updated versions of modules
 2019-08-24 12:57:29 +04:00
Auroir
c4e68ef539 Formatted Model Summary (#348) 
* Formatted Model Summary

Aligns the model summary output using f-string formatting. The logic structure of the base class has not been changed, only the lines put into `model_summary_text`. Output width is calculated from keys & values and will scale to show a clean summary for any model/platform.

GPU VRAM has been added as an output. Incorrect detection of VRAM is possible in broken environments and GPUs of different sizes can report the same name. Showing it here adds clarity for the user and for issue tickets.

Concatenation changed from "\r\n" to "\n", CRLF end of lines for Windows are handled transparently so using it here caused extra blank lines in the summary txt file.

**Examples:**
Using CUDA + SAE-LIAE
```
============= Model Summary ==============
==                                      ==
==         Model name: SAE              ==
==                                      ==
==  Current iteration: 16               ==
==                                      ==
==----------- Model Options ------------==
==                                      ==
==         batch_size: 4                ==
==        sort_by_yaw: False            ==
==        random_flip: True             ==
==         resolution: 128              ==
==          face_type: f                ==
==         learn_mask: True             ==
==     optimizer_mode: 1                ==
==              archi: liae             ==
==            ae_dims: 256              ==
==          e_ch_dims: 42               ==
==          d_ch_dims: 21               ==
== multiscale_decoder: False            ==
==         ca_weights: False            ==
==         pixel_loss: False            ==
==   face_style_power: 0.0              ==
==     bg_style_power: 0.0              ==
==    apply_random_ct: False            ==
==           clipgrad: False            ==
==                                      ==
==------------- Running On -------------==
==                                      ==
==       Device index: 0                ==
==               Name: GeForce GTX 1080 ==
==               VRAM: 8.00GB           ==
==                                      ==
==========================================
```
Colab
```
========== Model Summary ==========
==                               ==
==         Model name: SAE       ==
==                               ==
==  Current iteration: 39822     ==
==                               ==
==-------- Model Options --------==
==                               ==
==         batch_size: 24        ==
==        sort_by_yaw: True      ==
==        random_flip: False     ==
==         resolution: 128       ==
==          face_type: f         ==
==         learn_mask: True      ==
==     optimizer_mode: 2         ==
==              archi: liae      ==
==            ae_dims: 222       ==
==          e_ch_dims: 34        ==
==          d_ch_dims: 16        ==
== multiscale_decoder: True      ==
==         ca_weights: True      ==
==         pixel_loss: False     ==
==   face_style_power: 2.0       ==
==     bg_style_power: 1.5       ==
==    apply_random_ct: False     ==
==           clipgrad: True      ==
==                               ==
==--------- Running On ----------==
==                               ==
==       Device index: 0         ==
==               Name: Tesla K80 ==
==               VRAM: 11.00GB   ==
==                               ==
===================================
```
Using OpenCL + H128
```
=========================== Model Summary ===========================
==                                                                 ==
==        Model name: H128                                         ==
==                                                                 ==
== Current iteration: 0                                            ==
==                                                                 ==
==------------------------- Model Options -------------------------==
==                                                                 ==
==        batch_size: 4                                            ==
==       sort_by_yaw: False                                        ==
==       random_flip: True                                         ==
==        lighter_ae: False                                        ==
==        pixel_loss: False                                        ==
==                                                                 ==
==-------------------------- Running On ---------------------------==
==                                                                 ==
==      Device index: 0                                            ==
==              Name: Advanced Micro Devices, Inc. gfx900 (OpenCL) ==
==              VRAM: 7.98GB                                       ==
==                                                                 ==
=====================================================================
```
Using CPU (output trimmed)
```
==------- Running On --------==
==                           ==
==       Using device: CPU   ==
==                           ==
===============================
```
multi_gpu support is retained (output trimmed)
```
==------------- Running On -------------==
==                                      ==
==    Using multi_gpu: True             ==
==                                      ==
==       Device index: 1                ==
==               Name: Geforce GTX 1080 ==
==               VRAM: 8.00GB           ==
==       Device index: 2                ==
==               Name: Geforce GTX 1080 ==
==               VRAM: 8.00GB           ==
==                                      ==
==========================================
```

Low VRAM warning (output trimmed)
```
==------------- Running On -------------==
==                                      ==
==       Device index: 0                ==
==               Name: Geforce GTX 1050 ==
==               VRAM: 2.00GB           ==
==                                      ==
==========================================
/!\
/!\ WARNING:
/!\ You are using a GPU with 2GB or less VRAM. This may significantly reduce the quality of your result!
/!\ If training does not start, close all programs and try again.
/!\ Also you can disable Windows Aero Desktop to increase available VRAM.
/!\
```

* Fix indent
 2019-08-16 18:35:27 +04:00
iperov
e8c5f168bd Merge branch 'master' of https://github.com/iperov/DeepFaceLab 2019-08-11 11:17:38 +04:00
iperov
b72d5a3f9a fixed error "Failed to get convolution algorithm" on some systems 
fixed error "dll load failed" on some systems
Expanded eyebrows line of face masks. It does not affect mask of FAN-x converter mode.
 2019-08-11 11:17:22 +04:00
Josh Johnson
e2bc65d5f0 Fix issue with RTX GPU and TensorFlow (#322) 
An issue affecting at least 2070 and 2080 cards (possibly other RTX cards too) requires auto growth to be enabled for TensorFlow to work.

I don't know enough about the impact of this change to know whether this ought to be made optional or not, but for RTX owners, this simple change fixes TensorFlow errors when generating models.
 2019-08-02 16:40:41 +04:00
fakerdaker
582c974851 Colab choose random preview (#316) 2019-07-25 11:29:31 +04:00
iperov
8484060e01 Trainer: added option for all models 
Enable autobackup? (y/n ?:help skip:%s) :
Autobackup model files with preview every hour for last 15 hours. Latest backup located in model/<>_autobackups/01

SAE: added option only for CUDA builds:
Enable gradient clipping? (y/n, ?:help skip:%s) :
Gradient clipping reduces chance of model collapse, sacrificing speed of training.
 2019-06-20 10:42:55 +04:00
Jakob6174
ea1d59f620 Update ModelBase.py (#283) 
Typo: 'NotImplementeError' --> 'NotImplementedError'
 2019-06-19 13:02:19 +04:00
iperov
66a12a973a fix choosing preview image on options override 2019-05-14 17:00:25 +04:00
iperov
d6e8dde481 'sort by yaw' option now can be overriden each run 2019-05-14 09:33:53 +04:00
iperov
39809ff7dc upd 2019-05-11 16:15:16 +04:00
iperov
6143962248 SAE: option apply_random_ct now True by default 2019-05-10 19:50:19 +04:00
iperov
5a9498543d refactoring 2019-05-09 18:05:50 +04:00
iperov
d2011f8c32 upd doc 2019-05-08 09:32:32 +04:00
iperov
d8c1e2e58b change help for apply_random_ct option 2019-05-07 10:47:29 +04:00
iperov
a805f81142 SAE: added test option: 'Apply random color transfer to src faceset' 2019-05-06 11:34:56 +04:00
iperov
f26267fddf sae: upd help 2019-05-02 09:06:22 +04:00
iperov
54fa75aa13 fix 2019-05-02 00:39:18 +04:00
iperov
6dc1d43fe9 fix 2019-05-01 20:03:15 +04:00
iperov
9577c2d254 fix 2019-05-01 19:59:43 +04:00
iperov
2a8dd788dc SAE: added option 'Pretrain the model?', 
Pretrain the model with large amount of various faces. This technique may help to train the fake with overly different face shapes and light conditions of src/dst data. Face will be look more like a morphed. To reduce the morph effect, some model files will be initialized but not be updated after pretrain: LIAE: inter_AB.h5 DF: both decoders.h5. The longer you pretrain the model the more morphed face will look. After that, save and run the training again.
 2019-05-01 19:55:27 +04:00
iperov
659aa5705a _ 2019-04-30 17:14:02 +04:00
iperov
efb22ecce0 5 secs for override settings for colab 2019-04-30 07:32:15 +04:00
iperov
2cd028fb67 _ 2019-04-28 21:16:27 +04:00
iperov
836693d64f added option 'Choose image for the preview history? (y/n skip:' 2019-04-27 10:03:54 +04:00
iperov
9e05e58ac4 _ 2019-04-26 21:43:43 +04:00
iperov
0e088f6415 _ 2019-04-25 14:46:39 +04:00
iperov
9535a657d2 SAE: revert back CA weights option 2019-04-24 13:44:12 +04:00
iperov
b6711b97a3 SAE added CA weights 2019-04-24 11:38:45 +04:00
iperov
947feac047 refactorings 2019-04-24 11:28:58 +04:00
iperov
e1da9c56b4
SAE collapse fix (#245) 
* test

* _

* _

* upd dev_poseest

* SAE: finally collapses are fixed

* fix batch size help
 2019-04-24 09:38:26 +04:00
iperov
d62785ca5a _ 2019-04-24 00:36:09 +04:00
iperov
747e39145e _ 2019-04-23 18:34:00 +04:00
iperov
ea7ec74721 change help for batch size 2019-04-23 18:31:12 +04:00
iperov
385d5dabbf remove df-s and liae-s 2019-04-23 18:28:17 +04:00
iperov
2809d495c2 _ 2019-04-23 17:31:58 +04:00
iperov
268b402513 _ 2019-04-23 17:03:53 +04:00
iperov
47f9bad42b SAE: df-s liae-s archis 2019-04-23 17:03:30 +04:00
iperov
8969e80275 _ 2019-04-23 14:07:38 +04:00
iperov
565c131b97 SAE: removed option 'remove gray border', because it makes the model very resource intensive 2019-04-23 08:31:13 +04:00
iperov
e58197ca22 initial code to extract umdfaces.io dataset and train pose estimator 2019-04-23 08:14:09 +04:00
iperov
51a917facc back to 100* loss 2019-04-22 10:42:17 +04:00
iperov
cd2767cca0 SAE: back to batchnormalization 2019-04-22 10:28:01 +04:00