snapshot...

modules/deeplearning/cnn_cld_frac.py

@@ -61,33 +61,17 @@ label_idx = params.index(label_param)
 print('data_params: ', data_params)
 print('label_param: ', label_param)
 
-# x_134 = np.arange(134)
-# y_134 = np.arange(134)
-# x_64 = np.arange(64)
-# y_64 = np.arange(64)
-# x_134_2 = x_134[3:131:2]
-# y_134_2 = y_134[3:131:2]
-# t = np.arange(0, 64, 0.5)
-# s = np.arange(0, 64, 0.5)
-#
-# x_128_2 = x_134[3:131:2]
-# y_128_2 = y_134[3:131:2]
-# x_128 = x_134[3:131]
-# y_128 = y_134[3:131]
-
-#----------- New
-# x_134_2 = x_134[1:134:2]
-# t = np.arange(1, 66, 0.5)
-#--------------------------
-
-x_64 = np.arange(64)
-y_64 = np.arange(64)
-t = np.arange(0, 64, 0.5)
-s = np.arange(0, 64, 0.5)
-x_128_2 = slice(3, 131, 2)
-y_128_2 = slice(3, 131, 2)
-x_128 = slice(3, 131)
-y_128 = slice(3, 131)
+
+x_138 = np.arange(138)
+y_138 = np.arange(138)
+
+slc_x_2 = slice(0, 138, 2)
+slc_y_2 = slice(0, 138, 2)
+slc_x = slice(1, 67)
+slc_y = slice(1, 67)
+
+lbl_slc_x = slice(4, 132)
+lbl_slc_y = slice(4, 132)
 
 
 def build_residual_conv2d_block(conv, num_filters, block_name, activation=tf.nn.relu, padding='SAME',
@@ -268,19 +252,22 @@ class CNN:
         data_norm = []
         for param in data_params:
             idx = params.index(param)
-            tmp = input_data[:, idx, y_128, x_128]
+            tmp = input_data[:, idx, slc_y_2, slc_x_2]
+            tmp = tmp[:, slc_y, slc_x]
             tmp = normalize(tmp, param, mean_std_dct, add_noise=add_noise, noise_scale=noise_scale)
             # tmp = resample_2d_linear(x_64, y_64, tmp, t, s)
             data_norm.append(tmp)
         # --------------------------
         param = 'refl_0_65um_nom'
         idx = params.index(param)
-        tmp = input_data[:, idx, y_128, x_128]
+        tmp = input_data[:, idx, slc_y_2, slc_x_2]
+        tmp = tmp[:, slc_y, slc_x]
         tmp = normalize(tmp, param, mean_std_dct, add_noise=add_noise, noise_scale=noise_scale)
         # tmp = resample_2d_linear(x_64, y_64, tmp, t, s)
         data_norm.append(tmp)
         # --------
-        tmp = input_data[:, label_idx, y_128, x_128]
+        tmp = input_data[:, label_idx, slc_y_2, slc_x_2]
+        tmp = tmp[:, slc_y, slc_x]
         tmp = np.where(np.isnan(tmp), 0, tmp)  # shouldn't need this
         data_norm.append(tmp)
         # ---------
@@ -288,7 +275,7 @@ class CNN:
         data = data.astype(np.float32)
         # -----------------------------------------------------
         # -----------------------------------------------------
-        label = input_data[:, label_idx, y_128, x_128]
+        label = input_data[:, label_idx, lbl_slc_y, lbl_slc_x]
         label = self.get_label_data(label)
         label = np.expand_dims(label, axis=3)
 
@@ -444,15 +431,15 @@ class CNN:
 
         input_2d = self.inputs[0]
         print('input: ', input_2d.shape)
-        # conv = tf.keras.layers.Conv2D(num_filters, kernel_size=5, strides=1, padding='VALID', activation=None)(input_2d)
-        conv = input_2d
+        conv = tf.keras.layers.Conv2D(num_filters, kernel_size=3, strides=1, padding='VALID', activation=activation)(input_2d)
+        # conv = input_2d
         print('input: ', conv.shape)
 
-        conv = conv_b = tf.keras.layers.Conv2D(num_filters, kernel_size=2, strides=1, kernel_initializer='he_uniform', activation=activation, padding='SAME')(input_2d)
+        conv = conv_b = tf.keras.layers.Conv2D(num_filters, kernel_size=3, strides=1, kernel_initializer='he_uniform', activation=activation, padding='SAME')(conv)
 
-        conv = tf.keras.layers.Conv2D(num_filters, kernel_size=2, strides=1, kernel_initializer='he_uniform', activation=activation, padding='SAME')(conv)
+        # conv = tf.keras.layers.Conv2D(num_filters, kernel_size=3, strides=1, kernel_initializer='he_uniform', activation=activation, padding='SAME')(conv)
 
-        conv = tf.keras.layers.Conv2D(num_filters, kernel_size=2, strides=2, kernel_initializer='he_uniform', activation=activation, padding='SAME')(conv)
+        # conv = tf.keras.layers.Conv2D(num_filters, kernel_size=3, strides=2, kernel_initializer='he_uniform', activation=activation, padding='SAME')(conv)
         print(conv.shape)
 
         if NOISE_TRAINING: