Autoencoder for images in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the necessary library for building a neural network model.

Computer Vision

from tensorflow import [1]

Drag options to blanks, or click blank then click option'

Akeras

Bmatplotlib

Cnumpy

Dpandas

Attempts:

3 left

💡 Hint

Common Mistakes

Importing unrelated libraries like pandas or matplotlib.

Forgetting to import keras from tensorflow.

The keras module from TensorFlow is used to build neural network models including autoencoders.

2fill in blank

medium

Complete the code to define the input shape for the autoencoder model.

Computer Vision

input_img = keras.Input(shape=[1])

Drag options to blanks, or click blank then click option'

A(28, 28, 1)

B(100,)

C(3, 224, 224)

D(None,)

Attempts:

3 left

💡 Hint

Common Mistakes

Using a flat vector shape like (100,) instead of image dimensions.

Using color image shape (3, 224, 224) which is incorrect here.

The input shape for grayscale images of size 28x28 pixels is (28, 28, 1).

3fill in blank

hard

Fix the error in the code to compile the autoencoder model with the correct loss function.

Computer Vision

autoencoder.compile(optimizer='adam', loss='[1]')

Drag options to blanks, or click blank then click option'

Aprecision

Bmse

Ccategorical_crossentropy

Daccuracy

Attempts:

3 left

💡 Hint

Common Mistakes

Using classification losses like categorical_crossentropy.

Using metrics like accuracy or precision as loss.

For autoencoders reconstructing images, the mean squared error (mse) loss is commonly used.

4fill in blank

hard

Fill both blanks to create the encoder part of the autoencoder using convolution and pooling layers.

Computer Vision

x = keras.layers.Conv2D(32, (3, 3), activation='relu', padding='same')(input_img)
x = keras.layers.[1]((2, 2), padding='same')(x)

Drag options to blanks, or click blank then click option'

AFlatten

BDense

CDropout

DMaxPooling2D

Attempts:

3 left

💡 Hint

Common Mistakes

Using Dense or Flatten layers directly after Conv2D in encoder.

Using Dropout which is for regularization, not pooling.

The encoder uses Conv2D followed by MaxPooling2D to reduce image size while extracting features.

5fill in blank

hard

Fill all three blanks to build the decoder part with upsampling and convolution layers to reconstruct the image.

Computer Vision

x = keras.layers.[1]((2, 2))(encoded)
x = keras.layers.Conv2D(32, (3, 3), activation='relu', padding='[2]')(x)
decoded = keras.layers.Conv2D(1, (3, 3), activation='[3]', padding='same')(x)

Drag options to blanks, or click blank then click option'

AMaxPooling2D

Bsame

Csigmoid

DUpSampling2D

Attempts:

3 left

💡 Hint

Common Mistakes

Using MaxPooling2D in decoder which reduces size instead of increasing.

Using relu activation in output layer instead of sigmoid.

The decoder upsamples the encoded features using UpSampling2D, then applies convolution with same padding and a sigmoid activation to reconstruct the image pixels between 0 and 1.