sync remote changes

cgan_earth/code_files/networks.py

@@ -69,7 +69,6 @@ def forward(self, noise, labels, Training=True, ratio=2):
             for layer in self.gen:
                 x = layer(x)
             # upsample to give output spatial size (img_length, img_length)
-            print(torch.sigmoid(self.final_conv(x)).shape)
             return torch.sigmoid(self.final_conv(x))
 
 

cgan_earth/code_files/util.py

@@ -372,7 +372,10 @@ def change_noise(label, original_noise, netG, n_classes, z_dim=64, lf=4, device=
 
     lbl = test_label.repeat(1, 1, lf, max_len).to(device)
     imgs = np.array([])
-    noise = original_noise
+    if method == 'combined':
+        noise, bool_tensor = vary_noise(original_noise, value/3, ratio=0.5)
+    else:
+        noise = original_noise
     step = 0.0
     if step_size >= 1:
         num_img = 1
@@ -412,7 +415,7 @@ def change_noise(label, original_noise, netG, n_classes, z_dim=64, lf=4, device=
             elif method == 'sub':
                 noise = torch.add(noise, value)
             elif method == 'combined':
-                noise = vary_noise(noise, value/2, ratio=0.5)
+                noise, bool_tensor = vary_noise(noise, value/2, ratio=0.5, boolean=bool_tensor)
     return imgs, noise, netG
 
 def animate(path, imgs, fps=24):

cgan_earth/code_files/video.py

@@ -40,16 +40,20 @@ def replace_noise(original_noise, z_dim, lf, ratio, device):
     new_noise[:, :, :, -1] = torch.randn(1, z_dim, lf, device=device)
     return new_noise
 
-def vary_noise(original_noise, value, ratio):
+def vary_noise(original_noise, value, ratio, boolean=None):
     new_noise = torch.zeros_like(original_noise)
-    for idx0 in range(original_noise.shape[0]):
-        for idx1 in range(original_noise.shape[1]):
-            old = original_noise[idx0][idx1].clone().flatten()
-            old[torch.randint(len(old), (int(len(old)*ratio),))] = torch.add(old[torch.randint(len(old), (int(len(old)*ratio),))], value)
-            new_noise[idx0][idx1] = old.reshape(original_noise.shape[-2], original_noise.shape[-1])
-    boolean = torch.eq(original_noise, new_noise)
-    new_noise[boolean] = torch.sub(new_noise[boolean], value)
-    return new_noise
+    if boolean == None:
+        for idx0 in range(original_noise.shape[0]):
+            for idx1 in range(original_noise.shape[1]):
+                old = original_noise[idx0][idx1].clone().flatten()
+                old[torch.randint(len(old), (int(len(old)*ratio),))] = torch.add(old[torch.randint(len(old), (int(len(old)*ratio),))], value)
+                new_noise[idx0][idx1] = old.reshape(original_noise.shape[-2], original_noise.shape[-1])
+        boolean = torch.eq(original_noise, new_noise)
+        new_noise[boolean] = torch.sub(new_noise[boolean], value)
+    else:
+        new_noise[~boolean] = torch.add(original_noise[~boolean], value)
+        new_noise[boolean] = torch.sub(original_noise[boolean], value)
+    return new_noise, boolean
 
 def uniform_transit(label1_channel, label2_channel, cur_label, l_step_size):
     '''

cgan_earth/make_video.py

@@ -32,14 +32,14 @@
 sea_lbl = [3]
 snow_lbl = [4]
 lf = 16
-ratio = 2
+ratio = 4
 
 # test1: forest, then transit to sea, then roll in sea
 # the speed currently must start with the lowest possible speed to make sure the noise has right dimension
 imgs1, noise, netG = roll_video(proj_path, sea_lbl, netG(z_dim+n_classes, img_length), n_classes, z_dim, lf=lf, device=device, ratio=ratio, n_clips=5, step_size=0.25)
 imgs2, noise, netG = transit_video(sea_lbl, forest_lbl, n_classes, noise, netG, lf=lf, ratio=ratio, device=device, step_size=0.25, z_step_size=0.1, l_step_size=0.2, transit_mode='uniform')
 imgs3, noise, netG = roll_video(proj_path, forest_lbl, netG, n_classes, z_dim, lf=lf, device=device, ratio=ratio, n_clips=15, step_size=0.25, original_noise=noise)
-imgs4, noise, netG = change_noise(forest_lbl, noise, netG, n_classes, z_dim, lf=lf, device=device, ratio=ratio, n_clips=30, step_size=0.25, value=0.02, method='combined')
+imgs4, noise, netG = change_noise(forest_lbl, noise, netG, n_classes, z_dim, lf=lf, device=device, ratio=ratio, n_clips=30, step_size=0.25, value=0.00001, method='combined')
 
 # concatenante the imgs together and make video
 imgs = np.vstack((imgs1, imgs2))