Tutotrial 1 - Revision history

Ramanikeshav: /* Coding it Up! */

2018-05-31T22:48:21Z

Coding it Up!

2018-05-31T22:47:31Z

Coding it Up!

2018-05-31T22:46:12Z

Deep autoencoder for RGB images

2018-05-31T22:45:44Z

Sample Images

2018-05-31T22:45:10Z

Sample Image

2018-05-31T22:44:32Z

Deep autoencoder for RGB images

2018-03-07T10:54:28Z

Sample Image

2018-03-07T10:54:17Z

Sample Image

2018-03-07T10:53:58Z

Sample Image

2018-03-07T10:53:48Z

Sample Image

@@ Line 187: / Line 187: @@
 *Now that we've reduced a 784 pixel image into a 128 pixel image, we must build it up. For this purpose, we use the UpSampling2D function. Providing an argument of (2,2) doubles the value of the dimensions.<br /><PRE>x = Conv2D(8, (3, 3), activation='relu', padding='same')(encoded)&#10;x = UpSampling2D((2, 2))(x)&#10;x = Conv2D(8, (3, 3), activation='relu', padding='same')(x)&#10;x = UpSampling2D((2, 2))(x)&#10;x = Conv2D(16, (3, 3), activation='relu')(x)&#10;x = UpSampling2D((2, 2))(x)&#10;decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(x)</PRE>
 *It might be even surprising to note that only a minor modification is required for RGB images and that is to change the the last layer's(decoded layer) first argument to 3. Everything from here on,  follows from the corresponding deep autoencoder code.
-*Just as a final tip: Don't forget to import the Conv2D, MaxPooling2D, and UpSampling2D functions!
+*Just as a final tip: Don't forget to import the Conv2D, MaxPooling2D, and UpSampling2D functions and the input_img shape changes for RGB images!
 =Sample Images=

@@ Line 187: / Line 187: @@
 *Now that we've reduced a 784 pixel image into a 128 pixel image, we must build it up. For this purpose, we use the UpSampling2D function. Providing an argument of (2,2) doubles the value of the dimensions.<br /><PRE>x = Conv2D(8, (3, 3), activation='relu', padding='same')(encoded)&#10;x = UpSampling2D((2, 2))(x)&#10;x = Conv2D(8, (3, 3), activation='relu', padding='same')(x)&#10;x = UpSampling2D((2, 2))(x)&#10;x = Conv2D(16, (3, 3), activation='relu')(x)&#10;x = UpSampling2D((2, 2))(x)&#10;decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(x)</PRE>
 *It might be even surprising to note that only a minor modification is required for RGB images and that is to change the the last layer's(decoded layer) first argument to 3. Everything from here on,  follows from the corresponding deep autoencoder code.
 =Sample Images=

← Older revision		Revision as of 22:46, 31 May 2018
Line 161:		Line 161:

	*Change 3: While plotting the image, every occurrence of the shape <tt>(28, 28)</tt> must be replaced with the shape <tt>(32, 32, 3)</tt>. Furthermore, we must remove every occurrence of the <tt>plt.gray()</tt>.		*Change 3: While plotting the image, every occurrence of the shape <tt>(28, 28)</tt> must be replaced with the shape <tt>(32, 32, 3)</tt>. Furthermore, we must remove every occurrence of the <tt>plt.gray()</tt>.
−
−	~~And voila! Our autoencoder has learnt to encode and decode images!~~
−	~~Until the next tutorial, it's goodbye from the Knights who say 'ni'!~~

	=Convolutional autoencoder for grayscale & RGB images=		=Convolutional autoencoder for grayscale & RGB images=

@@ Line 192: / Line 192: @@
 =Sample Images=
-The following is a sample from the MNIST dataset. The top row is the actual image and the bottom row is the predicted image:
+The following is a sample from the MNIST dataset. The top row is the actual image and the bottom row is the predicted image (Deep AE):
 [[File:deep_autoencoder.jpg|frameless]]
-The following is a sample from the CIFAR 10 dataset. The top row is the actual image and the bottom row is the predicted image:
+The following is a sample from the CIFAR 10 dataset. The top row is the actual image and the bottom row is the predicted image (Deep AE):
 [[File:col_deep_autoenc.png|frameless]]

@@ Line 191: / Line 191: @@
 *It might be even surprising to note that only a minor modification is required for RGB images and that is to change the the last layer's(decoded layer) first argument to 3. Everything from here on,  follows from the corresponding deep autoencoder code.
-=Sample Image=
+=Sample Images=
 The following is a sample from the MNIST dataset. The top row is the actual image and the bottom row is the predicted image:
 [[File:deep_autoencoder.jpg|frameless]]

@@ Line 164: / Line 164: @@
 And voila! Our autoencoder has learnt to encode and decode images!
 Until the next tutorial, it's goodbye from the Knights who say 'ni'!
 =Sample Image=

@@ Line 168: / Line 168: @@
 The following is a sample from the MNIST dataset. The top row is the actual image and the bottom row is the predicted image:
 [[File:deep_autoencoder.jpg|frameless]]
-[[File:col_deep_autoenc.jpg]]
+[[File:col_deep_autoenc.png]]