Course Description
Students are introduced to deep learning fundamentals and architectures with a focus on computer vision tasks such as image classification, object detection, image segmentation, image synthesis, and 3D reconstruction. Relevant case studies will be discussed. Students will learn to build neural networks for computer vision problems using Python and Tensorflow. Upon completion of the course, students will be able to:
- Describe and implement basic image processing tasks such as filtering, edge and feature detection
- Develop computer vision applications using OpenCV and python
- Describe the deep learning fundamentals
- Understand convolutional neural networks and use it in high level computer vision applications such as image classification, recognition and object detection
- Use Tensorflow to design and develop neural networks for vision applications
Course Administration
Instructor
Dr. Jiju Poovvancheri (e-mail:jiju.poovvancheri@smu.ca)
Lectures
- Mondays (10:00am-11:15am) -LA 181
- Wednesdays (10:00am-11:15am)-LA 181
Recitations
- Wednesdays (11:30am-12:45pm)- SB 155
Office Hours (via MS Teams)
- Mondays (11:30 am -1:30 pm)
- Wednesdays (1:00 – 2:00 pm)
- Fridays (10:30 am-1:30 pm)
Course Pages
- MS Class Teams (Lecture slides, discussion forum, grades)
- Github (Set up & installation info, links to software libraries, starter codes of lab exercises & assignments)
Tentative Schedule (Fall 2022)
Date
Topic
Projects
Low level Vision
Sept. 7
Sept. 12
Sept. 14
Sept. 21
Sept. 28
Oct. 3
Oct. 5
Oct. 12
Introduction to the course
Image Formation
Spatial Filters
Non-linear Filters (Bilateral)
Image Gradients, Canny Edge
Feature Detection (Harris)
SIFT Descriptor, Matching
Mid term Test
Project 1: Hybrid Images
Project 2: Feature Matching
Deep Learning Fundamentals
Oct. 17
Oct. 19
Oct. 24
Oct. 26
Oct. 31
Nov. 2
Supervised Learning
Feed Forward NN
Gradient Descent and Variants
Backpropagation
Regularization, Training
Convolutional Neural networks
Project 3: Binary Classifier
High level Computer Vision
Nov. 14
Nov. 16
Nov. 21
Nov. 23
Nov. 28
Nov. 30
Dec. 5
Dec. 7
Image Classification
CNN Architectures
Object Detection -R-CNN
Face Detection (Siamese Networks)
Object Detection -YOLO
Segmentation
Point Clouds, 3D Reconstruction
Review
Project 4: Digit Recognition
Project 5: Face Recognition
Recitations (Tentative schedule)
- Recitations 1 – 2- Image Filtering (Hybrid Images)
- Recitations 3-5- Feature Matching
- Recitations 6-8- Binary Classifier
- Recitations 8-9-Digit Recognition (MNIST)
- Recitations 9-11- Face Recognition
Grading
Projects (40%)
- There will be five programming projects.
- The programming problems will involve python programming using OpenCV and Tensorflow libraries
- Each project carries 10% weightage.
- Lowest grade will be dropped
- Each project will be evaluated during the recitations
- The students will get two recitations to complete each project.
- Attendance in recitations is mandatory to get the project marks.
Midterm(20%)
- There will be one mid term exam (closed book). The tests will be conducted in the second week of October 2022
- Used to evaluate your knowledge of course contents.
- To be held during the lecture hour.
- If you miss the quiz for any reason: (1) You must contact your instructor within 48 hours, and (2) You will be required to fill out and submit a Declaration of extenuating Circumstances form.
Final Exam (40%)
- Saint Mary’s University ID is required.
- The final is open book with no electronics allowed.
- Scheduled by the Registrar during the formal exam period.
- Cumulative and will cover all material discussed in the course.
- You must pass the final exam to pass the course.
Notes:
The final mark will be a letter grade based on the scale described in Section 5 of the Academic Regulations in the University Calendar. There is no curving of grades, or grading based on rank (e.g., a certain number of “A” grades, etc.). Final grades are truncated to 1 decimal place and then rounded (.5 to .9 are rounded up, .0 to .4 are rounded down) to the nearest whole number. There will be no supplementary examinations.
Text Books
Required: Computer Vision: Algorithms and Applications, 2nd Edition
Richard Szelski
Wiley
Reference: Computer Vision: A Modern Approach, 2nd Edition,
David Forsyth and Jean Ponce
Pearsoned
Reference: Deep Learning
Ian Goodfellow, Yoshua Bengio and Aaron Courville
MIT Press