COMP 776: Computer Vision in 3D World

Spring 2025

Instructor: Roni Sengupta
Tuesday and Thursday, 3:30-4.45pm, SN 11

[ Overview | Course Details | Schedule | Pre-Requisite ]

Course Description

This is an advanced graduate course focusing on fundamentals and recent development in 3D Computer Vision. The course will teach basic first principles of 3D Computer Vision and then discuss how modern deep learning techniques and first principles can work together to produce various state-of-the-art 3D reconstruction and generation techniques. The course will be lecture based. However the students will be assigned papers for reading and they are expected to lead the discussion and answer detailed questions about the papers. nThis course will also involve working on a research topic centered around 3D Computer Vision.

Office Hours: By appointment only (SN 255).

Target Audience

MS & PhD students who have basic ideas in Deep Learning and Computer Vision, and interested in diving deep into the world of 3D Computer Vision

Goal/Student Learning Outcomes

(a) Develop fundamental mathematical models of 3D Computer Vision (b) Understand how modern deep learning techniques combined with fundamental concepts produce SOTA results (c) Develop hands-on coding skill in 3D Computer Vision through assignments and course projects (d) Learn how to read papers and discuss them (c) Improve presentation skills.

Grading

  • Course Project: 60% grade
  • 1 3D Math Assignments: 10% grade
  • Mid-term in-class exam: 20% grade
  • 1 Paper Presentation: 10% grade

Course Projects

Format: Course Projects have 2 tracks, you need to choose 1: (I) Research and development of novel 3D algorithms or applications (II) Implementation and analysis of existing 3D computer vision algorithms.

Track I Research

Goal is to write a research paper that can be submitted to Neurips, Siggraph Asia, AAAI, CVPR or other conferences. The topic of the paper should involve 3D Computer Vision.

Team: The course encourage students from slightly different research backgrounds to form a team and explore new exciting research projects. Maximum group size is 2. Group size of 1 can be permitted under special circumstances.

Deliverables: 5 mins project pitch, 15 mins mid-term presentation, 15 mins final presentation. You will need to write minimum 4 page paper in CVPR format.

Grading: Grading will be based on quality of research, presentation, and actual progress throughout the semester. Project Pitch: 5pts; Midterm: 20pts -> 10pts presentation + Q&A, 10pts actual progress; Final: 35pts -> 10pts presentation + Q&A, 5pts actual progress, 20 pts for paper quality (e.g. 16-20- CVPR submittable, 10-16 - CVPR workshop submittable, below 10 - not even workshop quality). You will need to clarify if the proposed project is a part of the research project you are working with your advisor or not. Expectations and grades will be scaled accordingly.

Track II Implementation and Analysis

You will be provided a list of 8 topics and you will be asked to choose 2 of these topics. For each topic you need to find 3 SOTA algorithms with publicly available code and a standard benchmark for evaluating the problem. After evaluating these 3 SOTA algorithms on standard benchmarks you need to capture your own images and analyse when these algorithms work well and when they fail.

Team: Individual projects are only allowed.

Deliverables: 5 mins project pitch highlighting which topics you will work on what algorithms+datasets you plan to use. Mid-term presentation will be on 1 topic and Final presentation is on the other topic. You need to prepare a technical report analysis your findings in CVPR format and submit your codebase for both projects independently.

Grading: Grading will be based on quality of analysis, presentation, and actual progress throughout the semester. Project Pitch: 5pts; Project I: 20pts -> 10pts presentation + Q&A, 10pts actual progress; Project II: 20pts -> 10pts presentation + Q&A, 10pts actual progress; 15pts for both tech reports.

Please read the guidelines for Track II PDF

Paper Reading & Presentation

Each student will be assigned only 1 paper. The instructor will provide the overview of the research topic and the overview of the paper. The student need to present detailed technical analysis of the paper for 15-20 minutes and answer questions. This presentation will be more discussion focused.

Misc.

Late Submissions: The class is structured around a tight paper presentation schedule. Therefore, late assignments will not be accepted. If you have other deadlines you can always reschedule your paper presentation ahead of time.

Academic Integrity: For your presentations and projects, you are allowed to use materials from external sources. However, you must clearly acknowledge those sources.

Exact paper names can be updated given this is an evolving field and new papers are coming out everyday. Exact paper names will be finalized 2 weeks before the presentation.

Schedule (Subject to Change)