946688567d
enhanced SAE model. You should to restart training. new default style power = 2.0 fix DF default batch sizes. upd readme
iperov2019-01-17 21:41:40 +04:00
1f2b1481ef
now you can train models on multiple GPU's on same workspace without cloning any folders. Model files names will be prefixed with GPU index if GPU choosed explicitly on train/convert start. if you leave GPU idx choice default, then best GPU idx will be choosed and model file names will not contain index prefix. It gives you possibility to train same fake with various models or options on multiple GPUs.
iperov2019-01-14 10:48:23 +04:00
bf5282f7ec
if you have multi GPU in a system, you can choose what GPU to use on model start
iperov2019-01-10 19:22:54 +04:00
48d0123f0b
SAE: new options face_style_power, bg_style_power instead of style_power. Zero - means dont use style. SAE: new option 'lighter_encoder'. Now model settings can be overrided by pressing enter in 2 seconds while model loading (works on Windows). Removed all MultiGPU models, because keras multi_gpu in fact doesn't work.
iperov2019-01-10 18:28:12 +04:00
7b70e7eec1
added new model U-net Face Morpher. removed AVATAR - useless model was just for demo removed MIAEF128 - use UFM insted removed LIAEF128YAW - use model option sort by yaw on start for any model All models now ask some options on start. Session options (such as target epoch, batch_size, write_preview_history etc) can be overrided by special command arg. Converter now always ask options and no more support to define options via command line. fix bug when ConverterMasked always used not predicted mask. SampleGenerator now always generate samples with replicated border, exclude mask samples. refactorings
iperov2019-01-02 17:26:12 +04:00
64537204bb
Sorry, no more linux/mac/docker support. If someone want to support it - create linux/mac fork, I will reference them in readme.
iperov2019-01-01 19:48:03 +04:00
ffddfeabdb
added converter --seamless-erode-mask-modifier . This mask used to define area for opencv seamless cloning function. So if you erode it little bit , you can decrease flickering effect on some frames.
iperov2018-12-20 13:50:38 +04:00
bc5ca1ab59
removed sort by hist-blur
iperov2018-12-20 12:57:16 +04:00
9926dc626a
refactorings, improved sort by hist-dissim
iperov2018-12-20 12:43:00 +04:00
b2d28e0b95
added converter option --use-predicted-mask . By default converter uses model's predicted mask, but now you can use DST mask.
iperov2018-12-19 23:28:48 +04:00
26571eff6a
added sort -by black . Places images which contains black area at end of folder. Useful to get rid of src faces which cutted by screen.
iperov2018-12-18 18:08:04 +04:00
db142bcc02
increased speed of 'sort by hist-dissim' by running in parallel
iperov2018-12-18 17:51:54 +04:00
43da6f84d6
improved sort by blur. Now it's better than laplacian, and multiprocessed.
iperov2018-12-18 16:51:36 +04:00
iperov