snapshot...

modules/deeplearning/srcnn_l1b_l2.py

@@ -36,6 +36,7 @@ NOISE_TRAINING = False
 NOISE_STDDEV = 0.01
 DO_AUGMENT = True
 
+DO_SMOOTH = False
 DO_ZERO_OUT = False
 DO_ESPCN = False  # Note: If True, cannot do mixed resolution input fields (Adjust accordingly below)
 
@@ -267,7 +268,6 @@ class SRCNN:
             tmp = input_data[:, idx, :, :]
             tmp = tmp.copy()
             tmp = np.where(np.isnan(tmp), 0, tmp)
-            # tmp = smooth_2d(tmp, sigma=1.0)
             if DO_ESPCN:
                 tmp = tmp[:, slc_y_2, slc_x_2]
             else:  # Half res upsampled to full res:
@@ -282,7 +282,6 @@ class SRCNN:
             tmp = input_data[:, idx, :, :]
             tmp = tmp.copy()
             tmp = np.where(np.isnan(tmp), 0, tmp)
-            # tmp = smooth_2d(tmp, sigma=1.0)
             # Full res:
             tmp = tmp[:, slc_y, slc_x]
             tmp = normalize(tmp, param, mean_std_dct)
@@ -293,7 +292,8 @@ class SRCNN:
         tmp = input_data[:, label_idx, :, :]
         tmp = tmp.copy()
         tmp = np.where(np.isnan(tmp), 0, tmp)
-        # tmp = smooth_2d(tmp, sigma=1.0)
+        if DO_SMOOTH:
+            tmp = smooth_2d(tmp, sigma=0.5)
         if DO_ESPCN:
             tmp = tmp[:, slc_y_2, slc_x_2]
         else:  # Half res upsampled to full res:
@@ -315,8 +315,9 @@ class SRCNN:
         # -----------------------------------------------------
         label = input_data[:, label_idx, :, :]
         label = label.copy()
-        # label = np.where(np.isnan(label), 0, label)
-        # label = smooth_2d(label, sigma=1.0)
+        if DO_SMOOTH:
+            label = np.where(np.isnan(label), 0, label)
+            label = smooth_2d(label, sigma=0.5)
         label = label[:, y_128, x_128]
 
         if label_param != 'cloud_probability':