COMP 790-175
Advanced Visual Computing: Physics-Informed AI
Spring 2026
COMP 790-175
COMP 790-175
Course Description
This graduate-level seminar-style course explores recent advances in visual computing systems, with a strong emphasis on physics-inspired AI. Topics include:
- •Classical inverse problems, forward models
- •Unrolled networks and Physics-inspired Neural Networks (PINNs) for imaging
- •Neural rendering and implicit neural representations
- •Diffractive neural networks
While the course emphasizes theoretical foundations, we will also connect these ideas to emerging developments in generative AI and their applications in visual computing. We will devote the first few classes to introductory lectures. The remainder of the semester will focus on student-led paper presentations and discussions. Students will also undertake a semester-long project resulting in a comprehensive review report.
Target Audience
Masters and PhD students interested in physics-informed AI/ML, inverse and differentiable rendering, graphics, imaging systems and visual computing.
Learning Objectives & Outcomes
By the end of this course, you will be able to:
- •Critically read, analyze, and discuss research papers
- •Improve technical presentation and communication skills
- •Understand the core concepts of computational imaging systems
- •Model and simulate optical/imaging systems
- •Develop mathematical models for light transport and visual computing
- •Gain hands-on programming and research skills through projects
Course Logistics
Meeting Times
Days: Monday/Wednesday
Time: 3:35pm - 4:50pm
Location: FB141
Instructor
Prof. Praneeth Chakravarthula
Email: praneeth@cs.unc.edu
Office Hours: By appointment only
(SN 205 or Zoom)
Prerequisites
Students are expected to have background knowledge in:
- •Linear algebra: vectors, matrices, tensors, dimensional analysis
- •Signal processing: convolutions, Fourier transforms, linear systems
- •Basic optics: lenses, light as rays and waves, cameras, image formation
- •Basic AI/ML: MLPs, CNNs, Transformers, generative models
- •Programming: Python (preferred), PyTorch/TensorFlow, MATLAB
Before taking the class, you should be comfortable with basic camera image formation, basics of AI/ML, and be able to read a recent ML conference paper and understand it at a conceptual level.
Grading
Seminar Format
Each student will present 1-2 papers during the semester.
Before Each Class
All students must:
- •Read the assigned papers
- •Submit a written review (format provided by the instructor)
Class Structure
- 1.The instructor provides an overview of the research topic and paper
- 2.The assigned student presents a 20-minute detailed technical analysis, followed by discussion and Q&A
- 3.Students are divided into two groups, arguing for acceptance vs. arguing for rejection of the paper, based on its merits and limitations
Note: Active participation is expected in discussions and debates. Attendance will be recorded for each session.
Study Material
- •Slides and additional readings will be posted on Canvas and/or course website
- •During the first 3 weeks, students are expected to watch and review video lecture series on PIML
- •High level overview of PIML and ACM lecture style on PIML provide quick background
Schedule & Topics
| Week | Date | Topic |
|---|---|---|
| Foundations | ||
| 1 | Jan 7, Wed | Introduction and fast forward |
| 2 | Jan 12, Mon | Linear algebra recap |
| 2 | Jan 14, Wed | Image recovery and inverse problems |
| 3 | Jan 19, Mon | No class (MLK Day) |
| 3 | Jan 21, Wed | Neural networks and diffusion models |
| Model-based Deep Learning and Unrolled Optimization | ||
| 4 | Jan 26, Mon | Deep Tensor ADMM-Net for Snapshot Compressive Imaging |
| 4 | Jan 28, Wed | End-to-End Optimization of Optics and Image Processing |
| Neural Scene Representations | ||
| 5 | Feb 2, Mon | NeRF Basics |
| 5 | Feb 4, Wed | Implicit Surfaces via Volume Rendering |
| Physics-augmented Neural Fields | ||
| 6 | Feb 9, Mon | No class (Well Being day) |
| 6 | Feb 11, Wed | Project proposals due |
| 7 | Feb 16, Mon | Continuum-aware NeRF (PAC-NeRF) |
| 7 | Feb 18, Wed | NeRF in Scattering Media |
| Differentiable Optics | ||
| 8 | Feb 23, Mon | Lens Design with Differentiable Ray Tracing |
| 8 | Feb 25, Wed | Hybrid Lens Design with Differentiable Wave Optics |
| Optical Neural Networks | ||
| 9 | Mar 2, Mon | Diffractive Deep Neural Networks |
| 9 | Mar 4, Wed | Spatially Varying Nanophotonic Neural Networks |
| Gaussian Splatting and Physics-aware Scene Models | ||
| 10 | Mar 9, Mon | 3D Gaussian Splatting |
| 10 | Mar 11, Wed | Physics Integrated Gaussians |
| Spring Break | ||
| 11 | Mar 16, Mon | No Class (Spring Break) |
| 11 | Mar 18, Wed | No Class (Spring Break) |
| Graph Neural Networks | ||
| 12 | Mar 23, Mon | Intro to Graph Neural Networks |
| 12 | Mar 25, Wed | Interaction Networks for Learning Physics |
| Physics-inspired World Models and Simulators | ||
| 13 | Mar 30, Mon | GNNs as Learnable Physics Engines |
| 13 | Apr 1, Wed | Graph-based Physics Simulators |
| Physics-regularized Learning and Generative Priors | ||
| 14 | Apr 6, Mon | Deep Image Prior |
| 14 | Apr 8, Wed | GNNs and Generative Priors for Solving Inverse Problems |
| 15 | Apr 13, Mon | Invertible Generative Models |
| 15 | Apr 15, Wed | Diffusion Posterior Sampling |
| 16 | Apr 20, Mon | Buffer class, final report discussion |
| 16 | Apr 22, Wed | Buffer class, final report discussion |
| End of semester | ||
| 17 | Apr 27 | Buffer class, final report discussion |
* Schedule subject to change. Updated schedule will be posted on Canvas.
Final Project
All students in the class will write a “mini-paper” as a final project. The paper should extend one or more papers we covered in the class. Students should write code and carry out additional experiments and then write up the results in a standard conference paper format.
Group Work
- •Students are welcome to work in groups on the final project
- •Groups of two are expected to put twice as much work into the project, and similarly for larger groups
- •Groups with five or more people require special permission
- •Groups must include a “contributions” paragraph listing each author's contributions
Paper Length
Maximum paper length: 3 + n_students pages (not including references or contributions paragraph), where n_students is the number of people in the group.
Proposal Deadline
Feb 11th - All groups must submit a project proposal (single-paragraph description of experiments, datasets, methods, etc.)
Final Presentation
Groups will present during the final exam period. All students in each group are required to present some material.
Course Policies
Attendance & Late Work
- •Missing class without completing the corresponding assignment results in a zero for that session
- •If you miss a class where you are presenting, you must create the presentation beforehand and find someone else to present for you
- •If you miss a class in a non-presenting role, complete the assignment and send it before class starts
- •Late work for readings cannot be accepted since everyone must read the same papers at the same time
- •Final projects cannot be accepted after the scheduled final exam slot
Honor Code
All students are expected to follow the guidelines of the UNC honor code. It is particularly important that you cite the source of different ideas, facts, or methods and do not claim someone else's work as your own. If you are unsure about which actions violate the honor code, feel free to ask.
Accessibility
The University of North Carolina at Chapel Hill facilitates the implementation of reasonable accommodations for students with disabilities, chronic medical conditions, a temporary disability or pregnancy complications. Accommodations are determined through the Office of Accessibility Resources and Service (ARS). See the ARS Website for contact information or email ars@unc.edu.
Mental Health Resources
CAPS is strongly committed to addressing the mental health needs of a diverse student body through timely access to consultation and connection to clinically appropriate services. Visit their website at caps.unc.edu or visit their facilities on the third floor of the Campus Health Services building for a walk-in evaluation.
Title IX & Support Resources
Any student who is impacted by discrimination, harassment, interpersonal violence, sexual violence, sexual exploitation, or stalking is encouraged to seek resources on campus or in the community. Contact the Director of Title IX Compliance (Adrienne Allison – Adrienne.allison@unc.edu), Report and Response Coordinators (reportandresponse@unc.edu), Counseling and Psychological Services (confidential), or the Gender Violence Services Coordinators (gvsc@unc.edu; confidential). Additional resources are available at safe.unc.edu.