You are on page 1of 2

Unit-2-Convolutional Neural Networks

Part A
1. Differentiate object recognition and object detection.
2. Determine the convolution layer output shape if the input image shape is 128x128x3 and
it uses 32 filters of size 7x7 with a stride of 1 and padding of 2.
3. Determine the number of parameters if the input image shape is 128x128x3 and the
convolution layer uses 32 filters of size 7x7 with a stride of 1 and padding of 2.
4. List the advantages and limitations of data augmentation.
5. List the Deep Transfer Learning Strategies.

Part B
1. Explain the Convolution layer with TensorFlow implementation.
2. The MNIST dataset consists of 70000 28x28 grayscale digit images in 10 classes. There
are 60000 training images and 10000 test images. Develop the TensorFlow
implementation for the following.
i) Load the MNIST dataset
ii) Create a sequential model with the appropriate number of neurons in the output layer,
activation function, and loss function
iii) Fit the model with training data and validation data.
3. Explain the following with examples
(i) Padding
(ii) Striding
(iii) Pooling
4. Implement transfer learning by using VGG as the base model to classify MNIST
dataset.
i) Load the MNIST dataset
ii) create the base model using VGG
iii) Create a sequential model with the appropriate number of neurons in the output layer,
activation function, and loss function
iv) Fit the model with training data and validation data.

5. Develop the TensorFlow implementation of the model with the input shape of (512,512,1)
6. Explain data augmentation and its techniques with examples.

7. The CIFAR-10 dataset consists of 60000 32x32 color images in 10 classes, with 6000
images per class. There are 50000 training images and 10000 test images. Develop the
TensorFlow implementation for the following.
i) Load the CIFAR-10 dataset
ii) Create a sequential model with the appropriate number of neurons in the output layer,
activation function, and loss function
iii) Fit the model with training data and validation data.

8. Implement transfer learning by using VGG as the base model to classify CIFAR-10
dataset.
i) Load the CIFAR-10 dataset
ii) create the base model using VGG
iii) Create a sequential model with the appropriate number of neurons in the output layer,
activation function, and loss function
iv) Fit the model with training data and validation data.

You might also like