-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmodulation_CUGAN.yml
44 lines (38 loc) · 1.52 KB
/
modulation_CUGAN.yml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
name: Test_CUGAN_Range-Conds
suffix: ~ # add suffix to saved images
model: srgan
distortion: sr
scale: 1
gpu_ids: [0]
# dataset
datasets:
val0:
name: CBSD68_DoubleD_Blur10_Noise30
mode: LQGT # LQGT (for paired data) | LQ (for real-world data)
dataroot_GT: dataset/CBSD68/CBSD68
dataroot_LQ: dataset/CBSD68/blur0_noise30
# Important! Range for condition.
cond_init: [ 0.0, 0.0 ] # starting point of modulation
range_mode: 3 # which element you want to iterate, 0 for deblurring only, 1 for denoising only, 3 for both.
range_stride: 0.1 # stride for modulation: 0.1, 0.05, 0.01...
val1:
name: NIND
mode: LQ
dataroot_LQ: dataset/NIND
# Important! Range for condition.
cond_init: [ 0.0, 0.0 ] # starting point of modulation
range_mode: 3 # which element you want to iterate, 0 for deblurring only, 1 for denoising only, 2 for both.
range_stride: 0.1 # stride for modulation: 0.1, 0.05, 0.01...
network_G:
which_model_G: CUGAN
in_nc: 3 # channel of input
out_nc: 3 # channel of output
cond_dim: 2 # modulation dimension 2
stages_blocks_num : [2, 2, 2] # number of Modulated Residual Block (MRB) in every stage
stages_channels : [64, 128, 256] # channels of Modulated Residual Block (MRB) in every stage
downSample_Ksize: 2 # kernel size used for downsample
#### path
path:
pretrain_model_G: pretrain_models/CUGAN.pth
strict_load: true
evaluate_metric: ['psnr', 'psnr_y', 'ssim', 'ssim_y', 'lpips', 'dists']