UCSD CSE 272: Advanced Image Synthesis (Winter 2023)
Images we will render during the course. 3D data courtesy of Wenzel Jakob, Jonas Pilo, and Bernhard Vogl.
Course Description
This course discusses modern physically-based rendering techniques. Given a 3D scene description including the geometry, how surfaces and volumes reflect lights, the light source emission profiles, and the pose of a camera, physically-based rendering simulates the interactions between photons, surfaces, and volumes and produces an image. Physically-based rendering is central to computer graphics, and is becoming ever more crucial to domains outside of graphics such as computer vision, computational imaging, machine learning, and robotics, with applications in autonomous driving, training artificial intelligence agents, biomedical imaging, photography, and more. We will go through how we model the appearance of scenes (e.g., how do hair reflect lights? do objects change appearance when they become wet?), how we simulate light transport of surfaces and volumes efficiently, and how we invert the light transport process via differentiation.
Throughout the course, we will build a renderer with the capability of rendering layered materials, hair, volumes, caustics, with modern rendering algorithms. If you have taken
CSE 168 and want more -- you should come! If not, no worries, we will try to cover the basics at the beginning.
Required Knowledge
Logistics
Lectures: Monday/Wednesday/Friday 11:00am-11:50am Pacific time. The class will be on Zoom. See the Canvas site for the Zoom link.
Instructor office hour: Friday 3pm in my Zoom. See the Canvas site for the link.
TA office hour: Thursday 1pm in his Zoom. See the Canvas site for the link.
We will do most of the online discussions on the Discord channel. Please see the Canvas site for the channel link.
Grading
There will be 3 programming homeworks and 1 final project, each 25%.
Late penalty: score * clamp(1 - (seconds passed after midnight of the deadline day) / (86400*7), 0, 1)
We will use the time on Canvas to determine how many seconds have passed.
Homeworks and Projects
The homeworks involve quite a bit of programming and can be tough for inexperienced. Start early and ask questions!
Many of them will be based on the lajolla renderer.
Final Project (25%): Rendering Competition (due 3/18)
For the final project, we will take inspiration from Stanford's CS 348b's
rendering competition: the goal is to render an image by extending lajolla (or your own renderer!). It will be graded by both artistic values and technical sophistication. You only need to achieve one of them (though it would be great if you achieve both!).
For people interested in inverse rendering: the final project can also involve using rendering techniques for recovering appearance or geometry -- then you can render an image of the object you reconstruct.
Collaboration policy: for the homeworks, you need to do it yourself (you are free to discuss between peers). For the final project,
you can have a team maximum of 2 people.
Schedule (tentative)
1/3/2022 (Mon): Introduction (Homework 0 out) [slides] [recordings]
Cosine-weighted hemisphere sampling. Next event estimation. Multiple importance sampling. Textures. Shading normals.
1/7/2022 (Fri): Bidirectional Reflectance/Scattering Distribution Functions [slides] [recordings]
1/10/2022 (Mon): Uber BSDF (Homework 1 out) [slides] [recordings]
1/12/2022 (Wed): Normal map filtering [slides] [recordings]
1/14/2022 (Fri): Layered BSDFs [slides] [recordings]
1/17/2022 (Mon): Martin Luther King's day, no class this day
1/19/2022 (Wed): Hair and Cloth [slides] [recordings]
1/21/2022 (Fri): Wave-based BSDFs [slides] [recordings]
1/24/2022 (Mon): Participating media (Homework 2 out) [slides] [recordings]
Radiative tranfer equation. Transmittance. Phase function. Rayleigh scattering.
1/26/2022 (Wed): Transmittance estimation and free-flight sampling [slides] [recordings]
Ray marching. Delta tracking. Ratio tracking. Null-scattering formulation.
1/28/2022 (Fri): Microflake theory [slides] [recordings]
1/31/2022 (Mon): Diffusion approximation (Homework 1 due) [slides] [recordings]
2/2/2022 (Wed): Single-scattering specialization [slides] [recordings]
2/4/2022 (Fri): Differentiable rendering 1 [slides] [recordings (I removed the Q&A part of the video since I mentioned personal things.)]
Automatic differentiation.
Edge sampling. Reynolds transport theorem.
2/7/2022 (Mon): Differentiable rendering 2 [slides] [recordings]
2/11/2022 (Fri): Stratification [slides] [recordings]
Jittered sampling. Blue-noise sampling. Low-discrepancy sequences. Frequency analysis.
2/14/2022 (Mon): Path-space and bidirectional path tracing (Homework 3 out) [slides] [recordings]
2/16/2022 (Wed): Photon mapping and its combination with bidirectional path tracing [slides] [recordings]
Photon mapping. Bias-variance analysis of density estimation. UPS/VCM. UPBP.
2/18/2022 (Fri): Metropolis light transport [slides] [recordings]
Markov chain Monte Carlo. Kelemen-style and Veach-style. Langevin/Hamiltonian Monte Carlo.
2/21/2022 (Mon): President day, no class this day. (Homework 2 due)
2/23/2022 (Wed): Rendering specular light paths [slides] [recordings]
2/25/2022 (Fri): Multiple importance sampling++ [slides] [recordings]
2/28/2022 (Mon): Many-light rendering [slides] [recordings]
3/2/2022 (Wed): Production rendering [slides] [recordings]
Micropolygons.
Ptex. Programmable shaders. Texture caches. Case studies: PRman, Manuka, Hyperion, and Arnold.
3/4/2022 (Fri): GPU architectures [slides] [recordings]
Rasterization/ray tracing. Architecture design. Coherency. Parallelism.
3/7/2022 (Mon): Nanite/real-time rendering (Homework 3 due) [slides] [recordings]
Streaming. Real-time rasterization.
3/9/2022 (Wed): Guest lecture by Steve Rotenberg [slides] [recordings]
Topics on wave optics simulation.
3/11/2022 (Fri): 100 weird tricks for your renderers [slides] [recordings]
Alias table. TaggedPointer. Roughness-based attenuation. Monte Carlo debiasing. Split-sum trick. etc.
3/18/2022 (Fri): final project due