質問編集履歴

2018/12/05 12:53

投稿

kktmkmmk

スコア10

test CHANGED Viewed

File without changes

test CHANGED Viewed

File without changes

タイトル

2018/12/05 12:53

投稿

kktmkmmk

スコア10

test CHANGED Viewed

	@@ -1 +1 @@
1	- 保存したモデルの再利用時のエラーについて
1	+ [pycharm]保存したモデルの再利用時のエラーについて[tensorflow]

test CHANGED Viewed

File without changes

コードの修正

2018/12/03 10:49

投稿

kktmkmmk

スコア10

test CHANGED Viewed

File without changes

test CHANGED Viewed

@@ -28,15 +28,13 @@
-# Load development/test data:
 dir_dev = r"data"
 n_im_dev = 90 # How many images to load
-# Load images and create motion-corrupted frequency space
-# No normalization or rotations:
 X_dev, Y_dev = load_images_from_folder(  # Load images for evaluating model
@@ -180,8 +178,6 @@
         reuse=tf.AUTO_REUSE,
-        #reuse=False,
         variables_collections=None,
         outputs_collections=None,
@@ -214,8 +210,6 @@
         reuse=tf.AUTO_REUSE,
-       # reuse=False,
         variables_collections=None,
         outputs_collections=None,
@@ -226,13 +220,13 @@
-    # Reshape output from FC layers into array of size (n_im, n_H0, n_W0, 1):
     FC_M = tf.reshape(FC2, [tf.shape(x)[0], tf.shape(x)[1], tf.shape(x)[2], 1])
-    # Retrieve the parameters from the dictionary "parameters":
     W1 = parameters['W1']
@@ -242,33 +236,29 @@
-    # CONV2D: filters W1, stride of 1, padding 'SAME'
-    # Input size (n_im, n_H0, n_W0, 1), output size (n_im, n_H0, n_W0, 64)
     Z1 = tf.nn.conv2d(FC_M, W1, strides=[1, 1, 1, 1], padding='SAME')
-    # RELU
     CONV1 = tf.nn.relu(Z1)
-    # CONV2D: filters W2, stride 1, padding 'SAME'
-    # Input size (n_im, n_H0, n_W0, 64), output size (n_im, n_H0, n_W0, 64)
     Z2 = tf.nn.conv2d(CONV1, W2, strides=[1, 1, 1, 1], padding='SAME')
-    # RELU
     CONV2 = tf.nn.relu(Z2)
-    # DE-CONV2D: filters W3, stride 1, padding 'SAME'
-    # Input size (n_im, n_H0, n_W0, 64), output size (n_im, n_H0, n_W0, 1)
     batch_size = tf.shape(x)[0]
@@ -306,31 +296,31 @@
-    # Create Placeholders
     X, Y = create_placeholders(n_H0, n_W0)
-    # Initialize parameters
     parameters = initialize_parameters()
-    # Build the forward propagation in the tf graph
     forward_propagation(X, parameters)
-    # Add ops to save and restore all the variables
     saver = tf.train.Saver()
-    # Start the session to compute the tf graph
     with tf.Session() as sess:
@@ -356,8 +346,6 @@
-# Reconstruct the image using trained model
 _, Y_recon = model(X_dev)
 print('Y_recon.shape = ', Y_recon.shape)
@@ -368,9 +356,7 @@
-# Visualize the images, their reconstruction using iFFT and using trained model
-# 4 images to visualize:
 im1 = 32
@@ -382,15 +368,13 @@
-# iFFT back to image from corrupted frequency space
-# Complex image from real and imaginary part
 X_dev_compl = X_dev[:, :, :, 0] + X_dev[:, :, :, 1] * 1j
-#iFFT
 X_iFFT0 = np.fft.ifft2(X_dev_compl[im1, :, :])
@@ -402,7 +386,7 @@
-# Magnitude of complex image
 X_iFFT_M1 = np.sqrt(np.power(X_iFFT0.real, 2)
@@ -422,10 +406,6 @@
-# SHOW
-# Show Y - input images
 plt.subplot(341), plt.imshow(Y_dev[im1, :, :], cmap='gray')
 plt.title('Y_dev1'), plt.xticks([]), plt.yticks([])
@@ -444,7 +424,7 @@
-# Show images reconstructed using iFFT
 plt.subplot(345), plt.imshow(X_iFFT_M1, cmap='gray')
@@ -464,7 +444,7 @@
-# Show images reconstructed using model
 plt.subplot(349), plt.imshow(Y_recon[im1, :, :], cmap='gray')