You are on page 1of 1

Nirma University

Institute of Technology
Class Test, Sep. - 2021
Master of Computer Application, Semester-III
3MCAD310 : Deep Learning
Roll/ Supervisor’s initial
Exam No. with date
Time: 1 Hours and 15 min. ________ Max. Marks: 35

Instructions: 1. Attempt all questions.


2. Figures to right indicate full marks.
3. Draw neat sketches and assume suitable data wherever necessary.

Que. 1 Do as directed: (11)


C1
BL3 i) Pooling is an important process in Deep neural network designing. Justify (5)
with suitable examples and diagrams.

BL2 ii) Why do we prefer Convolutional Neural networks (CNN) over Artificial (6)
Neural networks (ANN) for image processing task? Justify your answer with
suitable scenario.

Que. 2 Explain in brief multi-layer perceptron training algorithm. (8)


C1
BL2

Que. 3 If we give the input a 3-D image of dimension 32 X 32 X 3 to the (8)


C4, Convolutional Neural Network, then determine the following after each
BL4 layer:

i) Dimension of the intermediate o/p images after each layer.


ii) Number of hyperparameter to learns at each stage

Consider the layer architecture of CNN as below:

Layer-1: Filter Size – 5 X 5, Number of Filters – 10, Stride – 1, Padding – 0

Layer-2: Filter Size – 4 X 4, Number of Filters – 20, Stride – 2, Padding – 1

Que. 4 Define the Object classification, Object localization in context of (8)


C2,3, Convolutional Neural networks (CNN). Also explain in brief the block diagram
BL3 of the object localization operation in CNN.

----------------x-----------------

Page 1 of 1

You might also like