Upload 2 files

config_web.yml

@@ -0,0 +1,53 @@
+data:
+    image_size: 64
+    channels: 3
+    num_workers: 2
+    train_data_dir: '/data/JGW/hsguan/JGWNET/train/'  # path to directory of train data
+    test_data_dir: '/data/JGW/hsguan/JGWNET/selected/'  # path to directory of test data
+    #test_data_dir: '/home/dachuang/hsguan/JGWNET/data2/test2/'
+    #test_save_dir: '/home/dachuang/hsguan/JGWNET/selected'   # path to directory of saving restored data
+    test_save_dir: '/home/dachuang/hsguan/JGWNET/result90'
+    val_save_dir: '/data/JGW/hsguan/validation/'
+    grid_r: 16
+    conditional: True
+    tensorboard: '/home/dachuang/hsguan/JGWNET/90logs'
+
+model:
+    in_channels: 3
+    out_ch: 3
+    ch: 128
+    ch_mult: [1, 2, 3, 4]
+    num_res_blocks: 2
+    attn_resolutions: [16, ]
+    dropout: 0.0
+    ema_rate: 0.999
+    ema: True
+    resamp_with_conv: True
+
+diffusion:
+    beta_schedule: linear
+    beta_start: 0.0001
+    beta_end: 0.02
+    num_diffusion_timesteps: 1000
+
+training:
+    patch_n: 8
+    batch_size: 8
+    n_epochs: 1000
+    n_iters: 2000000
+    snapshot_freq: 20  # model save frequency
+    validation_freq: 10000
+    resume: './diffusion_model_new'  # path to pretrained model
+    seed: 61 # random seed
+
+sampling:
+    batch_size: 1
+    last_only: True
+    sampling_timesteps: 100
+
+optim:
+    weight_decay: 0.01
+    optimizer: "Adam"
+    lr: 0.0001
+    amsgrad: False
+    eps: 0.00000001

data_web.py

@@ -0,0 +1,15 @@
+import numpy as np
+import torchvision
+import PIL
+
+
+def web_input(image):
+    image_transforms = torchvision.transforms.Compose([torchvision.transforms.ToTensor()])
+    input_img = PIL.Image.fromarray(image)
+    input_img = input_img.resize((100, 100), PIL.Image.LANCZOS)
+    wd_new, ht_new = input_img.size
+    wd_new = int(16 * np.ceil(wd_new / 16.0))
+    ht_new = int(16 * np.ceil(ht_new / 16.0))
+    input_img = input_img.resize((wd_new, ht_new), PIL.Image.LANCZOS)
+    return image_transforms(input_img).unsqueeze(0)
+