snapshot...

modules/deeplearning/srcnn_cld_frac.py

@@ -37,7 +37,7 @@ NOISE_TRAINING = False
 NOISE_STDDEV = 0.01
 DO_AUGMENT = True
 
-DO_SMOOTH = True
+DO_SMOOTH = False
 SIGMA = 1.0
 DO_ZERO_OUT = False
 DO_ESPCN = False  # Note: If True, cannot do mixed resolution input fields (Adjust accordingly below)
@@ -62,7 +62,7 @@ IMG_DEPTH = 1
 # label_param = 'cld_opd_dcomp'
 label_param = 'cloud_probability'
 
-params = ['temp_11_0um_nom', 'temp_12_0um_nom', 'refl_0_65um_nom', label_param]
+params = ['temp_11_0um_nom', 'refl_0_65um_nom', label_param]
 data_params_half = ['temp_11_0um_nom']
 data_params_full = ['refl_0_65um_nom']
 
@@ -131,43 +131,6 @@ def build_residual_conv2d_block(conv, num_filters, block_name, activation=tf.nn.
     return conv
 
 
-def build_residual_block_conv2d_down2x(x_in, num_filters, activation, padding='SAME', drop_rate=0.5,
-                                do_drop_out=True, do_batch_norm=True):
-    skip = x_in
-
-    conv = tf.keras.layers.Conv2D(num_filters, kernel_size=3, strides=1, padding=padding, activation=activation)(x_in)
-    conv = tf.keras.layers.MaxPool2D(padding=padding)(conv)
-    if do_drop_out:
-        conv = tf.keras.layers.Dropout(drop_rate)(conv)
-    if do_batch_norm:
-        conv = tf.keras.layers.BatchNormalization()(conv)
-
-    conv = tf.keras.layers.Conv2D(num_filters, kernel_size=3, strides=1, padding=padding, activation=activation)(conv)
-    if do_drop_out:
-        conv = tf.keras.layers.Dropout(drop_rate)(conv)
-    if do_batch_norm:
-        conv = tf.keras.layers.BatchNormalization()(conv)
-
-    conv = tf.keras.layers.Conv2D(num_filters, kernel_size=3, strides=1, padding=padding, activation=activation)(conv)
-    if do_drop_out:
-        conv = tf.keras.layers.Dropout(drop_rate)(conv)
-    if do_batch_norm:
-        conv = tf.keras.layers.BatchNormalization()(conv)
-
-    skip = tf.keras.layers.Conv2D(num_filters, kernel_size=3, strides=1, padding=padding, activation=None)(skip)
-    skip = tf.keras.layers.MaxPool2D(padding=padding)(skip)
-    if do_drop_out:
-        skip = tf.keras.layers.Dropout(drop_rate)(skip)
-    if do_batch_norm:
-        skip = tf.keras.layers.BatchNormalization()(skip)
-
-    conv = conv + skip
-    conv = tf.keras.layers.LeakyReLU()(conv)
-    print(conv.shape)
-
-    return conv
-
-
 def upsample(tmp):
     tmp = tmp[:, slc_y_2, slc_x_2]
     tmp = resample_2d_linear(x_2, y_2, tmp, t, s)
@@ -395,10 +358,6 @@ class SRCNN:
         # -----------------------------------------------------
         label = input_data[:, label_idx, :, :]
         label = label.copy()
-        # if DO_SMOOTH:
-        #     label = np.where(np.isnan(label), 0, label)
-        #     label = smooth_2d(label, sigma=SIGMA)
-        #     # label = median_filter_2d(label)
         label = label[:, y_128, x_128]
         label = get_label_data(label)
 
@@ -518,16 +477,13 @@ class SRCNN:
 
         conv_b = build_residual_conv2d_block(conv_b, num_filters, 'Residual_Block_4', kernel_size=KERNEL_SIZE, scale=scale)
 
-        # conv_b = build_residual_conv2d_block(conv_b, num_filters, 'Residual_Block_5', kernel_size=KERNEL_SIZE, scale=scale)
-
-        # conv_b = build_residual_conv2d_block(conv_b, num_filters, 'Residual_Block_6', kernel_size=KERNEL_SIZE, scale=scale)
+        conv_b = build_residual_conv2d_block(conv_b, num_filters, 'Residual_Block_5', kernel_size=KERNEL_SIZE, scale=scale)
 
-        conv_b = build_residual_block_conv2d_down2x(conv_b, num_filters, activation)
+        conv_b = build_residual_conv2d_block(conv_b, num_filters, 'Residual_Block_6', kernel_size=KERNEL_SIZE, scale=scale)
 
         conv_b = tf.keras.layers.Conv2D(num_filters, kernel_size=3, strides=1, activation=activation, kernel_initializer='he_uniform', padding=padding)(conv_b)
 
-        # conv = conv + conv_b
-        conv = conv_b
+        conv = conv + conv_b
         print(conv.shape)
 
         if NumClasses == 2: