Computer Vision Course Notes
Chapter 1: Introduction
Welcome to the computer vision course
Instructor: Dimitris
Team: Three D Intelligence part of the computer vision lab
Focus Area
Study of machine perception, synthesis, and reasoning,
Emphasis on perception: synonym for computer vision
Course Objective
Understanding fundamentals of computer vision
Aims to develop mathematical models and algorithms to help computers decode and understand the surrounding world.
Key Concept: 3D Reconstruction
Input: Single color image
Output: 3D reconstruction of the depicted scene
Focus on human-centric scenarios, including interactions with physical objects.
Challenges in 3D Reconstruction
Color images contain color data but lack depth information.
Need to align 3D reconstructions of humans and objects.
Intuitive physics reasoning involved concerning physical contact (illustrated with heat maps showing likelihood of contact between objects).
Time Dimension
Processing static images and videos to reconstruct 3D models over time, effectively adding a fourth dimension.
Practical Applications
Applications in fields such as games, movies, and augmented reality.
For example, generating realistic hand movements for grasping objects.
Research Support
Supported by ELIS unit Amsterdam, Google, NVIDIA, and the European Research Council (ERC).
Course Format
Dimitris as the course tutor and coordinator
Emeeke as the DE4 (senior TA) for anonymous communication.
Team of TAs: Tyne, Lisa, Casper, Jorben, Jesse, Peter, and Maas.
Each TA will lead specific group sessions (Groups A to H).
Getting To Know Your Group
Resources available to find TA assignments and student groupings.
Emphasis on building strong relationships with TAs for an improved learning experience.
Course Statistics
Course run for the fourth year since 2023, with consistent changes in audience size and courses split.
Generally, course enrollments hover between 150-170 students.
Observed rate of students not needing to retake exams has improved, signifying better understanding and performance in the course.
Attendance Expectations
High encouragement for attendance in lectures and lab sessions (structured with alternating weekly pair formats).
Inclusion of WOO Club sessions for quizzes and knowledge recall with bonus grades associated.
Emphasis on working actively in lab sessions with the opportunity for peer collaboration.
Mandatory live grading sessions for understanding one’s submissions.
Chapter 2: Use The Canvas
Active Participation Importance
Engage to develop an analytical thinking style.
Course Tools
Datanose for scheduling.
Canvas for content retrieval and forum interactions (most crucial source).
Information Structure in Canvas
Canvas homepage filled initially with resources, later copied to course catalog and manual.
Access to speaking with course materials across multiple channels organized through Canvas.
Weekly Assignments
Required for all students, including those retaking the course.
Collaboration expected: pairs of two for each assignment to build soft skills.
Teams ideally formed from the same Learning Community (LC) groups to facilitate collaboration.
Deadline for forming teams set for end of Wednesday before the second lecture.
Submission Processes
Assignments to be submitted on ANS platform, with clear identification of team members required.
Key notes on code formatting and on correct submission processes when collaborating on projects for programming assignments.
Grading Structure
Average weekly grading method includes weight assignments:
Midterm and final exams each cover 50% of the course content and count for 40% of the overall grade.
Weekly assignments and live grading account for the remaining 15% each.
Requirement for passing grades includes maintaining a minimum score of 5.5 across all evaluation aspects.
Chapter 3: Exam Structures and Communication
Mock Exam
An example mock exam released to serve as a study guide and not a direct version of the final.
Midterm Exam Implementation
Request from students for the midterm exam has been addressed; will account for 50% of course material.
Tentatively scheduled after lecture-free weeks.
Live Grading Process
Mandatory attendance in two live grading sessions.
TAs assess understanding of assignments by engaging students in discussions about their work.
Grading Breakdown
Midterm Exam: covers 50% of material, accounts for 40% of grade.
Final Exam: covers remaining 50%, accounts for 40% of grade.
Weekly Assignments and Live Grading: each contribute 15% to final grade.
Points must maintain at least a minimum of 5.5 to pass.
Chapter 4: Small Sampling Probes
Introduction to Image Representation
Discussion on how computers encode and represent images for processing.
Introductory comparisons between human perceptions and machine interpretations of images.
Common Terms and Concepts
Definitions of terms: sampling grids, quantization, interpolation, and histograms.
Introduction of the role pixels play in the formation of images and the misconception of pixel shapes.
Methodology of Image Formation
Physical observation followed by light reflection captured by cameras versus human perception.
Summary of operations involved in converting natural objects to digital images through light measurements.
Key Mathematics
Formalizing the image process: functions defined mathematically as where each pixel's location maps onto a value measuring electromagnetic energy.
Conclusion of Chapter 4
Understanding the marriage of creative perception with computational representation, setting the foundation for deeper investigation of digital image processing.