Upload 2 files

x1_ITF_SkinDiffDetail_Lite_v1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94d368b633614958f84f335b129fd85abd30200e8fbc575b859ba6762116222b
+size 20099337

x1_ITF_SkinDiffDetail_Lite_v1.yml

@@ -0,0 +1,238 @@
+# python train.py -opt options/sr/x1_ITF_SkinDiffDetail_Lite_v1.yml
+name: x1_ITF_SkinDiffDetail_Lite_v1
+# the name that defines the experiment and the directory that will be created in the experiments directory.
+# name: debug_001_template  # use the "debug" or "debug_nochkp" prefix in the name to run a test session and check everything is working. Does validation and state saving every 8 iterations. Remove "debug" to run the real training session.
+use_tb_logger: false
+# wheter to enable Tensorboard logging or not. Output will be saved in: traiNNer/tb_logger/
+model: sr 
+# the model training strategy to be used. Depends on the type of model, from: https://github.com/victorca25/traiNNer/tree/master/codes/models
+scale: 1 # the scale factor that will be used for training for super-resolution cases. Default is "1".
+gpu_ids: [0] # the list of `CUDA_VISIBLE_DEVICES` that will be used during training, ie. for two GPUs, use [0, 1]. The batch size should be a multiple of the number of 'gpu_ids', since images will be distributed from the batch to each GPU.
+use_amp: true # select to use PyTorch's Automatic Mixed Precision package to train in low-precision FP16 mode (lowers VRAM requirements).
+use_swa: false # select to use Stochastic Weight Averaging
+use_cem: false # select to use CEM during training. https://github.com/victorca25/traiNNer/tree/master/codes/models/modules/architectures/CEM
+
+# Dataset options:
+datasets: # configure the datasets
+  train: # the stage the dataset will be used for (training)
+    name: x1_ITF_SkinDiffDetail_Lite_v1 # the name of your dataset (only informative)
+    mode: aligned 
+    # dataset mode: https://github.com/victorca25/traiNNer/tree/master/codes/data
+    dataroot_HR: [
+      #'K:/TRAINING/data/Skin_Diff2Nrml/hr_clean_tiles/'
+      '../datasets/Skin_DiffDetail/hr/'
+      ]
+    dataroot_LR: [
+      #'K:/TRAINING/data/Skin_Diff2Nrml/lr_clean_tiles/'
+      '../datasets/Skin_DiffDetail/lr_soft/'
+      ] # low resolution images
+    subset_file: null
+    use_shuffle: true
+    znorm: false
+    n_workers: 8
+    batch_size: 12
+    virtual_batch_size: 12
+    preprocess: crop
+    crop_size: 64
+    image_channels: 3
+
+    # Color space conversion
+    # color: 'y'
+    # color_LR: 'y'
+    # color_HR: 'y'
+    
+    # LR and HR modifiers.
+    # aug_downscale: 0.2
+    # shape_change: reshape_lr
+    
+    # Enable random downscaling of HR images (will fix LR pair to correct size)
+    hr_downscale: true
+    hr_downscale_types: [0, 3]
+    hr_downscale_amount: [1, 2, 4]
+    # #pre_crop: true
+    
+    # Presets and on the fly (OTF) augmentations
+    #augs_strategy: combo
+    #add_blur_preset: custom_blur
+    #add_resize_preset: custom_resize
+    #add_noise_preset: custom_noise
+    #aug_downscale: 0.2
+    resize_strat: pre
+    
+    # On the fly generation of LR:
+    # dataroot_kernels: 'KERNEL PATH !!!! CHANGE THIS OR COMMENT OUT'
+    #lr_downscale: false
+    #lr_downscale_types: ["linear", "bicubic", "nearest_aligned"]
+
+    # Rotations augmentations:
+    use_flip: true
+    use_rot: true
+    use_hrrot: true
+    
+    # Noise and blur augmentations:
+    #lr_blur: true
+    #lr_blur_types: {sinc: 0.2, iso: 0.2, ansio2: 0.4, sinc2: 0.2, clean: 3}
+    #noise_data: 'K:/TRAINING/traiNNer/noise_patches/'
+    #lr_noise: true
+    #lr_noise_types: {camera: 0.1, jpeg: 0.8, clean: 3}
+    #lr_noise2: false
+    #lr_noise_types2: {jpeg: 1, webp: 0, clean: 2, camera: 2}
+    #hr_noise: false
+    #hr_noise_types:  {gaussian: 1, clean: 4}
+    
+    # Color augmentations
+    # lr_fringes: false
+    # lr_fringes_chance: 0.4
+    # auto_levels: HR
+    # rand_auto_levels: 0.7
+    #lr_unsharp_mask: true
+    #lr_rand_unsharp: 0.7
+    # hr_unsharp_mask: true
+    # hr_rand_unsharp: 1
+    
+    # Augmentations for classification or (maybe) inpainting networks:
+    # lr_cutout: false
+    # lr_erasing: false
+
+  #val: 
+    #name: val_set14_part
+    #mode: aligned
+    #dataroot_B: '../datasets/val/hr'
+    #dataroot_A: '../datasets/val/lr'
+    
+    #znorm: false
+    
+    # Color space conversion:
+    # color: 'y'
+    # color_LR: 'y'
+    # color_HR: 'y'
+    
+
+path:
+    root: '../'
+    pretrain_model_G: '../experiments/pretrained_models/1x_DIV2K-Lite_SpongeBC1-Lite_interp.pth'
+    # pretrain_model_D: 'K:/TRAINING/data/models/x1_ITF_SkinDiff2Nrm_Lite_v3_208500_D.pth'
+    resume_state: '../experiments/x1_ITF_SkinDiffDetail_Lite_v1/training_state/latest.state'
+
+# Generator options:
+network_G: esrgan-lite # configurations for the Generator network
+
+
+# Discriminator options:
+network_D:
+    # ESRGAN (default)| PPON:
+    which_model_D: multiscale # discriminator_vgg_128 | discriminator_vgg | discriminator_vgg_128_fea (feature extraction) | patchgan | multiscale
+    norm_type: batch
+    act_type: leakyrelu
+    mode: CNA # CNA | NAC
+    nf: 32
+    in_nc: 3
+    nlayer: 3 # only for patchgan and multiscale
+    num_D: 3 # only for multiscale
+
+train:
+    # Optimizer options:
+    optim_G: adamp
+    optim_D: adamp
+    
+    # Schedulers options:
+    lr_scheme: MultiStepLR
+    lr_steps_rel: [50000, 100000, 200000, 300000]
+    lr_gamma: 0.5
+
+    # For SWA scheduler
+    swa_start_iter_rel: 0.05
+    swa_lr: 1e-4
+    swa_anneal_epochs: 10
+    swa_anneal_strategy: "cos"
+    
+    # Losses:
+    pixel_criterion: l1  # pixel (content) loss
+    pixel_weight: 0.05
+    feature_criterion: l1 # feature loss (VGG feature network)
+    feature_weight: 0.3
+    cx_type: contextual  # contextual loss
+    cx_weight: 1
+    cx_vgg_layers: {conv_3_2: 1, conv_4_2: 1}
+    #hfen_criterion: l1  # hfen
+    #hfen_weight: 1e-6 
+    #grad_type: grad-4d-l1  # image gradient loss
+    #grad_weight: 4e-1
+    # tv_type: normal  # total variation
+    # tv_weight: 1e-5
+    # tv_norm: 1
+    ssim_type: ssim  # structural similarity
+    ssim_weight: 0.05
+    lpips_weight: 0.25  # [.25] perceptual loss
+    lpips_type: net-lin
+    lpips_net: squeeze
+    
+    # Experimental losses
+    # spl_type: spl  # spatial profile loss
+    # spl_weight: 0.1
+    #of_type: overflow  # overflow loss
+    #of_weight: 0.1
+    # range_weight: 1  # range loss
+    # fft_type: fft  # FFT loss
+    # fft_weight: 0.2 #[.2]
+    color_criterion: color-l1cosinesim  # color consistency loss
+    color_weight: 0.1
+    # avg_criterion: avg-l1  # averaging downscale loss
+    # avg_weight: 5
+    # ms_criterion: multiscale-l1  # multi-scale pixel loss
+    # ms_weight: 1e-2
+    #fdpl_type: fdpl  # frequency domain-based perceptual loss
+    #fdpl_weight: 1e-3
+    
+    # Adversarial loss:
+    #gan_type: vanilla
+    #gan_weight: 4e-3
+    # freeze_loc: 4
+    # For wgan-gp:
+    # D_update_ratio: 1
+    # D_init_iters: 0
+    # gp_weigth: 10
+    # Feature matching (if using the discriminator_vgg_128_fea or discriminator_vgg_fea):
+    # gan_featmaps: true
+    # dis_feature_criterion: cb  # discriminator feature loss
+    # dis_feature_weight: 0.01
+    
+    # For PPON:
+    # p1_losses: [pix]
+    # p2_losses: [pix-multiscale, ms-ssim]
+    # p3_losses: [fea]
+    # ppon_stages: [1000, 2000]
+    
+    # Differentiable Augmentation for Data-Efficient GAN Training
+    # diffaug: true
+    # dapolicy: 'color,transl_zoom,flip,rotate,cutout'
+    
+    # Batch (Mixup) augmentations
+    #mixup: false
+    #mixopts: [blend, rgb, mixup, cutmix, cutmixup] # , "cutout", "cutblur"]
+    #mixprob: [1.0, 1.0, 1.0, 1.0, 1.0] #, 1.0, 1.0]
+    #mixalpha: [0.6, 1.0, 1.2, 0.7, 0.7] #, 0.001, 0.7]
+    #aux_mixprob: 1.0
+    #aux_mixalpha: 1.2
+   # mix_p: 1.2
+    
+    # Frequency Separator
+    #fs: true
+    #lpf_type: average
+    #hpf_type: average
+    
+    # Other training options:
+    manual_seed: 0
+    niter: 250000
+    # warmup_iter: -1
+    #val_freq: 5e3
+    # overwrite_val_imgs: true
+    # val_comparison: true
+    # metrics: 'psnr,ssim,lpips'
+    #grad_clip: auto
+    #grad_clip_value: 0.1 # "auto"
+
+logger:
+    print_freq: 50
+    save_checkpoint_freq: 500
+    overwrite_chkp: false