iperov
cd2767cca0
SAE: back to batchnormalization
2019-04-22 10:28:01 +04:00
iperov
3aef9a507b
fix
2019-04-22 00:52:16 +04:00
iperov
44aa530fda
multiple preview file support for colab
2019-04-22 00:49:50 +04:00
iperov
f54cc744bd
SAE fix model crashes
2019-04-22 00:40:23 +04:00
iperov
5e3bd75007
fix
2019-04-21 22:28:25 +04:00
iperov
16053c4ae6
SAE: revert back to batchnormalization
2019-04-21 22:17:41 +04:00
iperov
093ca4dba6
fix options in ModelBase.py
2019-04-21 21:08:45 +04:00
iperov
63beb3afd2
SAE: revert back df model from prev commit. LIAE should be restarted.
2019-04-21 08:31:05 +04:00
iperov
2cdf2745a2
SAE: model will not work with this update, you should to restart training,
...
applied paper https://arxiv.org/abs/1807.01442 https://github.com/aditya-grover/uae
2019-04-20 22:33:28 +04:00
iperov
7c5f762e78
SAE:fix multiscale decoder
2019-04-20 21:30:40 +04:00
iperov
8e16f4fa1b
fix
2019-04-20 16:10:58 +04:00
iperov
3ffdc50576
nothing interesting
2019-04-20 15:22:00 +04:00
iperov
046649e6be
update == 04.20.2019 == ( #242 )
...
* superb improved fanseg
* _
* _
* added FANseg extractor for src and dst faces to use it in training
* -
* _
* _
* update to 'partial' func
* _
* trained FANSeg_256_full_face.h5,
new experimental models: AVATAR, RecycleGAN
* _
* _
* _
* fix for TCC mode cards(tesla), was conflict with plaidML initialization.
* _
* update manuals
* _
2019-04-20 08:23:08 +04:00
iperov
58d7e990f4
refactorings, added motion blur to SampleProcessor for FANSegmentator trainer
2019-04-07 23:08:00 +04:00
iperov
5ac7e5d7f1
changed help message for pixel loss:
...
Pixel loss may help to enhance fine details and stabilize face color. Use it only if quality does not improve over time.
SAE:
previous SAE model will not work with this update.
Greatly decreased chance of model collapse.
Increased model accuracy.
Residual blocks now default and this option has been removed.
Improved 'learn mask'.
Added masked preview (switch by space key)
Converter:
fixed rct/lct in seamless mode
added mask mode (6) learned*FAN-prd*FAN-dst
added mask editor, its created for refining dataset for FANSeg model, and not for production, but you can spend your time and test it in regular fakes with face obstructions
2019-04-04 10:22:53 +04:00
iperov
01e98cde8e
upd pixel_loss help message
2019-03-31 19:21:04 +04:00
iperov
85c01e3b4a
Converter: added Apply super resolution? (y/n skip:n) : , Enhance details by applying DCSCN network.
...
refactorings
2019-03-28 21:50:27 +04:00
iperov
3cf3bb786e
rename samples to samplelib
2019-03-27 10:44:13 +04:00
iperov
070e938568
colab: final fix for preview filepath
2019-03-26 18:11:33 +04:00
iperov
e8a5a4bf10
_preview.jpg for colab
2019-03-26 17:58:29 +04:00
iperov
6241c8af41
preview file for colab
2019-03-26 17:14:59 +04:00
iperov
b03b147bae
refactoring
2019-03-26 11:09:44 +04:00
iperov
131b2b5c79
after save loss string now shows averaged value since last save
2019-03-25 18:14:04 +04:00
iperov
aa58d9e563
SAE: removed lightweight encoder
2019-03-25 10:05:30 +04:00
iperov
54239f8edf
sae: optimizer mode now can be overrided
2019-03-24 16:25:30 +04:00
iperov
37505d88e3
old SAE model will not work with this update.
...
Fixed bug when SAE can be collapsed during a time.
SAE: removed CA weights and encoder/decoder dims.
added new options:
Encoder dims per channel (21-85 ?:help skip:%d)
More encoder dims help to recognize more facial features, but require more VRAM. You can fine-tune model size to fit your GPU.
Decoder dims per channel (11-85 ?:help skip:%d)
More decoder dims help to get better details, but require more VRAM. You can fine-tune model size to fit your GPU.
Add residual blocks to decoder? (y/n, ?:help skip:n) :
These blocks help to get better details, but require more computing time.
Remove gray border? (y/n, ?:help skip:n) :
Removes gray border of predicted face, but requires more computing resources.
2019-03-24 15:35:02 +04:00
iperov
cbe2ebca7f
SAE: removed random normal initialization
2019-03-23 14:57:30 +04:00
iperov
97008291b8
fix
2019-03-20 14:43:02 +04:00
iperov
6169e6ba8a
fix fanseg
2019-03-20 09:08:42 +04:00
iperov
a3df04999c
removing trailing spaces
2019-03-19 23:53:27 +04:00
iperov
fa4e579b95
fix fanseg
2019-03-19 23:49:41 +04:00
iperov
034ad3cce5
upd fan segmentator
2019-03-19 19:44:14 +04:00
iperov
d71a310fd7
SAE: forgot to remove normalizing from tanh
2019-03-17 13:03:49 +04:00
iperov
1c8845d558
model fansegmentator for developing purposes
2019-03-16 20:57:51 +04:00
iperov
d6a45763a2
SAE: added option "Use CA weights":
...
Initialize network with 'Convolution Aware' weights. This may help to achieve a higher accuracy model, but consumes time at first run.
2019-03-16 12:54:36 +04:00
iperov
5076feb90f
SAE: revert back to sigmoid. Trainings must be restarted.
2019-03-16 09:01:11 +04:00
iperov
8da47fec13
fix ModelBase, nnlib
2019-03-13 20:53:59 +04:00
iperov
3375088669
SAE: optimizer_mode option now cannot be overridden by restart
2019-03-13 20:00:16 +04:00
iperov
a9026ccb67
fix ModelBase, nnlib
2019-03-13 19:50:16 +04:00
iperov
58763756f5
SAE: removed simple_optimizer . Added optimizer mode for tensorflow only (NVIDIA cards), allows to train x2-x3 bigger networks with normal Adam optimizer, consuming VRAM and CPU power.
2019-03-13 11:54:17 +04:00
iperov
3c6775bae7
fix ModelBase.py
2019-03-12 23:22:58 +04:00
iperov
97b6fabaab
change 'epoch' to 'iter',
...
added timestamp prefix to training string
2019-03-12 19:23:52 +04:00
iperov
46ff33bf89
SAE: dssim kernel size now depends on resolution
2019-03-12 09:49:40 +04:00
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
