 nagadomi
|
1900ac7500
Use PSNR for evaluation
|
il y a 9 ans |
|
nagadomi
|
664322fa97
Remove cleanup_model.lua; Change model format from binary to ascii
|
il y a 9 ans |
|
nagadomi
|
290a5f960b
Use Huber loss instead of MSE
|
il y a 9 ans |
|
nagadomi
|
9b238bd693
Use clearState()
|
il y a 9 ans |
|
nagadomi
|
9f935835dd
Add -save_history option
|
il y a 10 ans |
|
nagadomi
|
aaac6ed6e5
Refactor training loop
|
il y a 10 ans |
|
nagadomi
|
c72ec3112b
Add -random_unsharp_mask_rate option for photo
|
il y a 10 ans |
|
nagadomi
|
d8ba661d6b
Add -jpeg_chroma_subsampling_rate option for JPEG denoise training
|
il y a 10 ans |
|
nagadomi
|
42bd89151e
Add -gpu option in train.lua
|
il y a 10 ans |
|
nagadomi
|
5ab2c605c8
Fix header include order issues
|
il y a 10 ans |
|
nagadomi
|
797b45ae23
Use roundf-like clip for 8 bit-depth image
|
il y a 10 ans |
|
nagadomi
|
3ea16b3b86
tunable parameters
|
il y a 10 ans |
|
nagadomi
|
c773e18e59
Add trade-off parameter for noise reduction
|
il y a 10 ans |
|
nagadomi
|
903d945652
cleanup
|
il y a 10 ans |
|
nagadomi
|
b35a9ae7d7
tuning
|
il y a 10 ans |
|
nagadomi
|
490eb33a6b
Minimize the weighted huber loss instead of the weighted mean square error
|
il y a 10 ans |
|
nagadomi
|
da786e15ba
remove noise_scale training
|
il y a 10 ans |
|
nagadomi
|
3abc5a03e3
refactor
|
il y a 10 ans |
|
nagadomi
|
8dea362bed
sync from internal repo
|
il y a 10 ans |
|
nagadomi
|
54580ba8c0
add data augmentation method that uses overlay
|
il y a 10 ans |
|
nagadomi
|
e3d3a8355c
change training script
|
il y a 10 ans |
|
nagadomi
|
fd9dadd7a4
use settings.validation_crop
|
il y a 10 ans |
|
nagadomi
|
5b4d692f03
add support for RGB color space reconstruction
|
il y a 10 ans |
|
nagadomi
|
a706892b59
merge develop repo
|
il y a 10 ans |
|
nagadomi
|
1c1f2292d1
fix ugly chroma scaling
|
il y a 10 ans |
|
nagadomi
|
ffd52fdb2b
super typo fix
|
il y a 10 ans |
|
nagadomi
|
2231423056
update training script
|
il y a 10 ans |
|
nagadomi
|
1273b3609e
first commit
|
il y a 10 ans |
