snapshot...

modules/deeplearning/srcnn_l1b_l2.py

@@ -693,6 +693,7 @@ class SRCNN:
         pred = self.model([data], training=False)
         self.test_probs = pred
         pred = pred.numpy()
+        print('**: ', pred.shape, pred.min(), pred.max())
 
         return pred
 
@@ -774,19 +775,29 @@ def run_evaluate_static(in_file, out_file, ckpt_dir):
     # grd_c = gaussian_filter(grd_c, sigma=1.0)
     grd_c = grd_c[y_0:y_0+sub_y, x_0:x_0+sub_x]
     grd_c = grd_c.copy()
+    print(grd_c.shape)
     grd_c = np.where(np.isnan(grd_c), 0, grd_c)
     hr_grd_c = grd_c.copy()
     hr_grd_c = hr_grd_c[y_128, x_128]
+    print(hr_grd_c.shape)
     grd_c = grd_c[slc_y_2, slc_x_2]
+    print(grd_c.shape)
     grd_c = resample_2d_linear_one(x_2, y_2, grd_c, t, s)
+    print(grd_c.shape)
     grd_c = grd_c[y_k, x_k]
+    print(grd_c.shape)
     if label_param != 'cloud_probability':
         grd_c = normalize(grd_c, label_param, mean_std_dct)
+    print(grd_c.shape)
 
     # data = np.stack([grd_a, grd_b, grd_c], axis=2)
     #data = np.stack([grd_a, grd_c], axis=2)
     data = np.stack([grd_c], axis=2)
+    print(data.shape)
     data = np.expand_dims(data, axis=0)
+    print(data.shape)
+    dn = denormalize(grd_c, label_param, mean_std_dct)
+    return hr_grd_c, grd_c, dn
 
     nn = SRCNN()
     out_sr = nn.run_evaluate(data, ckpt_dir)