Tafila Technical University
Faculty of Engineering Total Score
Department of Computer Engineering
Computer Vision
Midterm Exam
April 28th, 2024
Student Name:
Instructor:
This exam consists of two questions. You must have five pages.
The time allowed for this exam is 60 minutes.
Question#1 Question#2
(18) (12)
Good Luck
Question#1
Choose the correct answer:
1] Computer vision is a discipline that studies how to reconstruct, interrupt and understand a 3d
scene from its ________.
A. 1d images
B. 2d images
C. 3d images
D. 4d images
2] In MLP, the bias is used to:
A. Give some inputs higher weights.
B. Add non-linearity to the prediction model.
C. Shift the activation function towards the positive or negative side.
D. Map any input value between 0 and 1.
3] Quantization of image would:
A. Reduce its dimensions.
B. Improve its quality
C. Reduce its gray level colors
D. Remove noise.
4] What is a typical function of the convolutional layers in a CNN regarding image features?
A. Pooling feature representations
B. Extracting and detecting features like edges and textures
C. Reducing the dimensionality of the input
D. Providing fully connected layers with data
5] In Gaussian filtering, what effect does increasing the value of sigma (σ) have on the output
image?
A. Increases the sharpness of edges
B. Increases the intensity of the image
C. Increases the blurring effect
D. Decreases the contrast of the image
6] What is the purpose of hysteresis thresholding in the Canny edge detector?
A. To eliminate noise
B. To connect edge pixels into continuous contours
C. To compute the gradient magnitude
D. To apply Gaussian smoothing
7] What is the purpose of image segmentation in computer vision?
A. Classifying images into categories
B. Identifying individual objects within an image
C. Extracting features from images
D. Enhancing image resolution
8] In an MLP with an input layer of 100 neurons, a hidden layer of 50 neurons, and an output
layer of 10 neurons, how many total parameters (weights and biases) are there if each neuron is
fully connected to the neurons in the adjacent layers?
A. 5,000
B. 5,100
C. 5,500
D. 5,560
9] What is the purpose of the activation function in a multi-layer perceptron (MLP)?
A. It defines the loss function used for training.
B. It adjusts the learning rate during backpropagation.
C. It introduces non-linearity to the network.
D. It initializes the weights and biases of the network.
10] In the context of evaluating edge detection algorithms, what does a precision value of 0.8
indicate?
A. The algorithm has detected 80% of the true edges.
B. 80% of the detected edges are true positives.
C. The algorithm has missed 20% of the true edges.
D. The ratio of true positive edges to the total number of ground truth edges is 0.8.
11] Which OpenCV function is commonly used to perform convolution-based image filtering?
A. cv::dilate()
B. cv::erode()
C. cv::filter2D()
D. cv::Canny()
12] In Gaussian filter, if we choose σ= 2, then, the best spatial filter size is:
A. 5x5 B. 9x9 C. 13x13 D. 17x17
Question#2
Given a 4-bit 5x5 digital image
1 11 12 13 9
3 4 10 9 14
6 7 5 9 10
5 13 5 8 10
13 10 13 14 12
Determine the following:
a. Sketch the histogram and determine whether the image is dark or bright?
b. Apply the following two masks to the image and determine the functionality of each one.
1. 3x3 Median Filter: Remove Noise
2. Detect Diagonal Edges
0 1 2
-1 0 1
-2 -1 0
3. Sharpen Image
-1 -1 -1
-1 9 -1
-1 -1 -1
