snapshot...

modules/deeplearning/cloud_opd_srcnn_viirs.py

@@ -70,18 +70,20 @@ N_X = N_Y = 1
 LEN_X = LEN_Y = 128
 
 if KERNEL_SIZE == 3:
-    slc_x = slice(2, N_X*LEN_X + 4)
-    slc_y = slice(2, N_Y*LEN_Y + 4)
-    slc_x_2 = slice(1, N_X*LEN_X + 6, 2)
-    slc_y_2 = slice(1, N_Y*LEN_Y + 6, 2)
+    slc_x_m = slice(1, int((N_X*LEN_X)/2) + 4)
+    slc_y_m = slice(1, int((N_Y*LEN_Y)/2) + 4)
+    slc_x = slice(3, N_X*LEN_X + 5)
+    slc_y = slice(3, N_Y*LEN_Y + 5)
+    slc_x_2 = slice(2, N_X*LEN_X + 7, 2)
+    slc_y_2 = slice(2, N_Y*LEN_Y + 7, 2)
     x_2 = np.arange(int((N_X*LEN_X)/2) + 3)
     y_2 = np.arange(int((N_Y*LEN_Y)/2) + 3)
     t = np.arange(0, int((N_X*LEN_X)/2) + 3, 0.5)
     s = np.arange(0, int((N_Y*LEN_Y)/2) + 3, 0.5)
     x_k = slice(1, N_X*LEN_X + 3)
     y_k = slice(1, N_Y*LEN_Y + 3)
-    x_128 = slice(3, N_X*LEN_X + 3)
-    y_128 = slice(3, N_Y*LEN_Y + 3)
+    x_128 = slice(4, N_X*LEN_X + 4)
+    y_128 = slice(4, N_Y*LEN_Y + 4)
 elif KERNEL_SIZE == 5:
     slc_x = slice(3, 135)
     slc_y = slice(3, 135)
@@ -120,7 +122,6 @@ def build_residual_conv2d_block(conv, num_filters, block_name, activation=tf.nn.
 
 
 def upsample(tmp):
-    tmp = tmp[:, slc_y_2, slc_x_2]
     tmp = resample_2d_linear(x_2, y_2, tmp, t, s)
     tmp = tmp[:, y_k, x_k]
     return tmp
@@ -290,9 +291,10 @@ class SRCNN:
 
         data_norm = []
         for param in data_params_half:
-            idx = params_i.index(param)
-            tmp = input_label[:, idx, :, :]
+            idx = params.index(param)
+            tmp = input_data[:, idx, :, :]
             tmp = np.where(np.isnan(tmp), 0, tmp)
+            tmp = tmp[:, slc_y_m, slc_x_m]
             tmp = upsample(tmp)
             tmp = normalize(tmp, param, mean_std_dct)
             data_norm.append(tmp)
@@ -302,23 +304,24 @@ class SRCNN:
             tmp = input_label[:, idx, :, :]
             tmp = np.where(np.isnan(tmp), 0, tmp)
 
-            lo, hi, std, avg = get_min_max_std(tmp)
-            lo = upsample_nearest(lo)
-            hi = upsample_nearest(hi)
-            avg = upsample_nearest(avg)
-            lo = normalize(lo, param, mean_std_dct)
-            hi = normalize(hi, param, mean_std_dct)
-            avg = normalize(avg, param, mean_std_dct)
+            # lo, hi, std, avg = get_min_max_std(tmp)
+            # lo = upsample_nearest(lo)
+            # hi = upsample_nearest(hi)
+            # avg = upsample_nearest(avg)
+            # lo = normalize(lo, param, mean_std_dct)
+            # hi = normalize(hi, param, mean_std_dct)
+            # avg = normalize(avg, param, mean_std_dct)
+            #
+            # data_norm.append(lo[:, slc_y, slc_x])
+            # data_norm.append(hi[:, slc_y, slc_x])
+            # data_norm.append(avg[:, slc_y, slc_x])
 
-            data_norm.append(lo[:, slc_y, slc_x])
-            data_norm.append(hi[:, slc_y, slc_x])
-            data_norm.append(avg[:, slc_y, slc_x])
-
-            # tmp = normalize(tmp, param, mean_std_dct)
-            # data_norm.append(tmp[:, slc_y, slc_x])
+            tmp = normalize(tmp, param, mean_std_dct)
+            data_norm.append(tmp[:, slc_y, slc_x])
         # ---------------------------------------------------
         tmp = input_label[:, label_idx_i, :, :]
         tmp = np.where(np.isnan(tmp), 0, tmp)
+        tmp = tmp[:, slc_y_2, slc_x_2]
         tmp = upsample(tmp)
         data_norm.append(tmp)
         # ---------