iperov
fd3b9add2f
SAE: added option "simple optimizer" allows to train bigger networks on same VRAM
...
nnlib: added DFLOptimizer is my own optimizer
2019-03-12 09:32:35 +04:00
iperov
3bad8dd8ec
fix
2019-03-11 21:53:28 +04:00
iperov
ee8dbcbc35
revert back Adam
2019-03-11 21:52:36 +04:00
iperov
e4637336ef
added ability to save optimizers states which work with K.function,
...
added custom Adam that can save 'iterations' param
2019-03-11 18:23:01 +04:00
iperov
714d4f0fca
small fix
2019-03-03 16:42:13 +04:00
iperov
31c2298b5f
Converter: added option for seamless to supress jittering,
...
Lenx,leny region now averaged by grayscale gradients,
now uses all CPU.
SAE: multiscale_decoder option default = False
update readme and manual_ru.pdf
2019-03-03 15:33:52 +04:00
iperov
438213e97c
manual extractor: increased FPS,
...
sort by final : now you can specify target number of images,
converter: fix seamless mask and exception,
huge refactoring
2019-02-28 11:56:31 +04:00
iperov
65752e044b
SAE: added support 64-256 resolution multiple of 16
2019-02-21 20:44:55 +04:00
iperov
a858732b1d
refactoring
2019-02-21 20:06:04 +04:00
iperov
97685ce0ae
added saving model_summary.txt
2019-02-21 19:55:44 +04:00
iperov
f0a20b46d3
SAE: added new archi 'vg'
2019-02-21 17:53:59 +04:00
iperov
72ba6b103c
added support of AMD videocards
...
added Intel's plaidML backend to use OpenCL engine. Check new requirements.
smart choosing of backend in device.py
env var 'force_plaidML' can be choosed to forced using plaidML
all tf functions transferred to pure keras
MTCNN transferred to pure keras, but it works slow on plaidML (forced to CPU in this case)
default batch size for all models and VRAMs now 4, feel free to adjust it on your own
SAE: default style options now ZERO, because there are no best values for all scenes, set them on your own.
SAE: return back option pixel_loss, feel free to enable it on your own.
SAE: added option multiscale_decoder default is true, but you can disable it to get 100% same as H,DF,LIAEF model behaviour.
fix converter output to .png
added linux fork reference to doc/doc_build_and_repository_info.md
2019-02-19 17:33:12 +04:00
iperov
3a9d450281
little fix
2019-02-16 21:14:15 +04:00
iperov
75eb7468ee
ConverterMasked: better lct
2019-02-14 18:43:24 +04:00
iperov
2bd983703e
ConverterMasked: removed default transfercolor,
...
added Apply color transfer to predicted face - modes rct / lct
2019-02-13 16:27:57 +04:00
iperov
535041f7bb
upd
2019-02-12 10:55:41 +04:00
iperov
429e7e6aee
upd nnlib.py
2019-02-12 09:30:38 +04:00
iperov
6c944d8989
upd readme
2019-02-11 21:26:51 +04:00
iperov
4ffb74fb79
upd some help in models
2019-02-11 17:20:13 +04:00
iperov
470fb9287a
SAE: remove rare sample booster. DSSIM->MSE transition now between 5-20k epochs.
2019-02-11 17:18:01 +04:00
iperov
f8e63970d2
H64, H128, DF, LIAEF128: added pixel loss option.
2019-02-11 12:05:54 +04:00
iperov
af3dd59f67
SAE: remove multiscale decoder option, default - true
2019-02-11 10:36:02 +04:00
iperov
854ab11de3
fix bug with samples that were not clipped after tanh-untanh transformations, upd README.md
2019-02-10 10:45:51 +04:00
iperov
51a13c90d1
SAE: you have to restart training,
...
added multiscale decoder as option.
mask now training as not multiscaled
2019-02-09 20:33:26 +04:00
iperov
4d37fd62cd
fix DFLJPG,
...
SAE: added "rare sample booster"
SAE: pixel loss replaced to smooth transition from DSSIM to PixelLoss in 15k epochs by default
2019-02-09 18:53:37 +04:00
iperov
ea7c65c4e5
upd ModelBase.py
2019-02-07 22:38:31 +04:00
iperov
a65b557b0e
loss history with epoch now saves to preview history images
2019-02-07 22:34:51 +04:00
iperov
ea7a2456c5
upgrading to numpy==1.16.1, fix ConverterMasked
2019-02-07 11:27:43 +04:00
iperov
0c1638e168
fix strange opencv behaviour with border_transparent
2019-02-07 11:14:41 +04:00
iperov
5a1fb199f4
added support of non-english characters in paths
2019-02-05 19:33:55 +04:00
iperov
c3b6330284
SAE: now asking face_type after resolution
2019-02-05 15:00:27 +04:00
iperov
883da4b9ed
fix Util.py, SAE: lowest ae dims now can be 32 (for testing)
2019-02-05 14:43:54 +04:00
iperov
241ae83e0b
upd readme
2019-02-05 13:27:18 +04:00
iperov
707313eb47
SAE+ConverterMasked now offer default values such as mode, erode, blur, values based on model settings.
2019-02-03 21:44:30 +04:00
iperov
e226ab5385
SAE: increased speed of training by 10-18%,
...
increased clipping border mask in full face mode results better transition of cheeks,
default archi now 'df'
2019-02-03 20:32:46 +04:00
iperov
541d6e1cc9
nothing interesting
2019-01-26 06:50:52 +04:00
iperov
e5b14672ee
SAE: upd help for new 'pixel loss' option
2019-01-25 10:02:30 +04:00
iperov
ef3cf392c5
SAE: added pixel loss option.
2019-01-25 09:56:40 +04:00
iperov
1e7a0836f7
nothing interesting
2019-01-24 21:28:54 +04:00
iperov
7f4d261a99
change SAE help message
2019-01-22 18:31:00 +04:00
iperov
6d7c56309d
SAE: removed adaptive kernel size. No backward compatibility.
2019-01-22 13:30:38 +04:00
iperov
b6c4171ea1
optimizations of nnlib and SampleGeneratorFace,
...
refactorings
2019-01-22 11:52:04 +04:00
iperov
fe66b6b2a1
SAE: added 'learn mask' option, default:Y,
...
SAE: added clipping mask at borders to remove artifact lines.
2019-01-19 16:15:49 +04:00
iperov
dd10d963d1
SAE: upd default params, upd readme
2019-01-19 08:58:25 +04:00
iperov
946688567d
enhanced SAE model. You should to restart training.
...
new default style power = 2.0
fix DF default batch sizes.
upd readme
2019-01-17 21:41:40 +04:00
iperov
22401cecfc
fix H128
2019-01-15 17:52:54 +04:00
iperov
ba06a71fff
SAE: changed default style power to 10.0 . Now style power is floating number in valid range 0.0 to 100.0
2019-01-14 21:24:33 +04:00
iperov
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.
H64 and H128: now you can choose 'Lighter autoencoder'. It is same as vram gb <= 4 before this update.
added archived_models.zip contains old experiments
RecycleGAN: archived
devicelib: if your system has no NVML installed (some old cards), then it will work with gpu_idx=0 as 'Generic GeForce GPU' with 2GB vram.
refactorings
2019-01-14 10:48:23 +04:00
iperov
b68612e6cd
SAE: 'adaptive kernel size' now as an option
2019-01-13 10:03:01 +04:00
iperov
46aa61050a
improvement SAE model. New option added. You should to restart training.
2019-01-12 10:10:50 +04:00
