Scribd
Upload
11 views2 pages

Sample Final Question 3

The document outlines the final exam for the course CSE 4883 Digital Image Processing at United International University, consisting of four questions worth a total of 40 marks. It covers various topics including traditional and machine learning segmentation techniques, pixel adjacency, morphological image processing, image restoration, and evaluation metrics for object detection models. Examinees are required to answer all questions and adhere to academic integrity rules.

Uploaded by

kawsarnewazchowdhury
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
11 views2 pages

Sample Final Question 3

The document outlines the final exam for the course CSE 4883 Digital Image Processing at United International University, consisting of four questions worth a total of 40 marks. It covers various topics including traditional and machine learning segmentation techniques, pixel adjacency, morphological image processing, image restoration, and evaluation metrics for object detection models. Examinees are required to answer all questions and adhere to academic integrity rules.

Uploaded by

kawsarnewazchowdhury
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

United International University (UIU)

Dept. of Computer Science and Engineering (CSE)


Final Exam, Spring 2024
Course: CSE 4883 Digital Image Processing
Total Marks: 40, Time: 2 hours
Any examinee found adopting unfair means will be expelled from the trimester/program as per UIU disciplinary rules.
There are FOUR questions. Answer all of them. Figures in the right-hand margin indicate full marks.

Take a look at the image of the bus above. Now, answer questions 1 a-c.

1a. If we want to find the outline of the bus, what type of traditional segmentation technique 3
would be best? Give justification is support of your choice.

1b. Now, if we were to use ML based (UNET) segmentation here, how would we proceed? Explain 4
the process starting from data annotation to the final model. Explain with necessary diagrams
for UNET.

1c Differentiate Among semantic, instance and panoptic segmentation. 3

2a 4*2
=8

Take a look at the 3x3 pixel grid given above. Now, find if the following pixel pairs are
m-adjacent or not. Consider that here V = {1}. Also find the Chessboard Distance, euclidean
distance and city-block distance between the pixel pairs.
I. e and c.
II. a and h
III. e and g.
IV. b and e.

2b Define region and boundary of a picture with an example pixel grid or a sample picture. 2

1 of 2
Take a look at the picture of a fly as given above, with a prominently green background only.
Npw answer Q 3a-3c.

3a Suggest ways we can remove the background from the above picture using either segmentation 2
or morphological image processing.

3b. If we want to find only the shape of the bug (without the feet extrusions). How may we do so 2+2
using morphological image processing? Explain with necessary diagrams.
Then, find how we may only find the outline of the bug using MIP.

3c Now, suppose that the picture above originally had salt and pepper noise in it? How may we 4
remove it using image restoration? Explain with the help of flowchart of the image restoration
process.

4a. 6

Take a look at the above confusion matrix of an object detection model that detects three
popular fruits from a picture. Now calculate the following for the classes “apple” and “banana”.
I. Accuracy
II. Precision
III. Recall
IV. F1 Score

Then, find only the macro precision and recall for the entire set.

4b. Based on the results from Q 4a, comment on how we may improve the overall accuracy of 4
classes apple and banana, based on the precision, recall, and F1 score.

2 of 2

You might also like