Carter Sifferman

Is it possible to reconstruct 3D geometry from measurements of a miniature time-of-flight proximity sensor? Yes! I'm excited to share our CVPR 2024 paper, "Towards 3D Vision with Low-Cost Single-Photon Cameras" 🧵thread below

@sidartsoma Great work, congrats Sid!

I’m excited to share our Nature paper on seeing around corners with consumer LiDARs! We show that consumer LiDAR sensors — in your smartphones, AR headsets, self-driving cars, and robots — can be used to see objects hidden around corners. youtube.com/watch?v=N3LEhh…

@jon_barron I agree on principle, but how is a grad student going to pay for that much API usage? I don’t know of any grad programs that provide credits / Claude code or similar to their grad students.

If I was a grad student today, I would: 1) Not write papers, 2) push my (agent-written) code to a public repo ~weekly, 3) maintain (via agents) a writeup.tex (manually verified) and a skill.md in the repo, and 4) work towards establishing skill usage as the new "citation" format.

Project Page: cpsiff.github.io/recovering_par… ArXiv: arxiv.org/pdf/2509.16132 YouTube: youtu.be/p6G4_JU5y2k Authors: Carter Sifferman*, Yiquan Li*, Yiming Li, Fangzhou Mu, Michael Gleicher, Mohit Gupta, Yin Li.

This approach has applications anywhere that benefits from low-cost, lightweight sensing, like wearables, mobile devices, and robotics. We hope this work will inspire further development in this direction!

📜New paper: Recovering Parametric Scenes from Very Few Time-of-Flight Pixels. To appear at ICCV 2025. We recover 3D parametric scenes (e.g. 6D object pose, human hand pose) from very few (8-15) distributed 3D pixels. 🧵thread below

@ukmlv Congrats and welcome!

Friends, I have a major update! After 8 wonderful years at WashU ESE/CSE, my team and I will be moving to UW-Madison ECE. I’m excited to collaborate with amazing new colleagues at UW, and will truly miss WashU and St. Louis—places where I’ve built many great memories. 🦡❤️

@zillyzindz And I caught you doing it

@anagh_malik Well deserved, congrats!

📣📣📣 Neural Inverse Rendering from Propagating Light 💡 just won Best Student Paper award at #CVPR!!!

Announcing the first #ICCP2025 Summer School at @UofT, July 19–20, 2025! Open to undergrad and grad students worldwide—join us for hands-on courses in computational imaging & photography. There are limited spaces available. Appication review begins: June 3, 23:59 ET 1/2

@dopenessmaxxer @jayparsons I was too x.com/CartSiff/statu…

@jayparsons I’m so curious as to which one of those dots in the graph is Austin, TX

I was once a skeptic of rental housing "filtering" -- the theory that new "luxury" apartments pull up higher-income renters from moderate-priced rentals, thereby benefitting moderate-income renters, and on down the line. I am no longer a skeptic. I became a believer once I realized several mistakes I made in the analysis -- the same mistakes I see some of my peers making now. Here are those common mistakes among filtering deniers: 1) You must focus on higher-supplied neighborhoods. Macro analysis doesn't work nearly as well. The more you build, the more clearly you see "filtering" play out. When you include low-supply neighborhoods in the analysis, you're obscuring the real story. 2) You cannot rely on the industry's asset class definitions based on non-rent measures like amenities or building age. You need to more specifically look at apartments by rent level, which is the foundation of "filtering." It's broadly correlated with classes, but not the same thing. 3) You can't assume "more supply" = "high supply." Very common mistake. In the 2010s decade, as example, we just didn't build that many apartments -- even in the places we *think* we did. For example, Texas probably built more apartments than any other state and yet rents outpaced national average-- making it harder to detect impact of filtering (even if still present in form of rents growing less than they otherwise would have absent the new supply.) It's not just supply. It's supply relative to demand. And in the 2010s, supply generally didn't keep pace with demand. The 2023-24 period shows us more clearly what happens when you build truly A LOT of new apartments. Rents fall -- even among the lowest-rent properties. (See chart below.) 4) You got to talk to people with eyes/ears on the ground -- especially those operating apartments at differing rent levels in high-supplied submarkets. They'll help you better understand what to look for in the data. 5) You need to be open to admitting you were wrong.

@jayparsons I was dying to know which dot was which so I did some searching, and found this plot with some labels. Hunstville, AL! I thought Austin would be further right, but given it's so large already it makes sense. source: dallasfed.org/-/media/docume…

@Michael_J_Black Is the code available yet?

Just in time for your #CVPR2025 evaluation! CameraHMR is the new state-of-the-art in parametric 3D human pose and shape (HPS) estimation and will appear at #3DV2025. There are 4 key contributions that make it so accurate and robust. 👇

@ZGojcic Big fan of NFL and other works from your team. Applied and emailed :)

Our team at NVIDIA is recruiting PhD research interns for next year. If you are interested in neural reconstruction, generative models, graphics and related topics. Please apply at the link below or contact me directly. :) nvidia.wd5.myworkdayjobs.com/en-US/NVIDIAEx…

@johnnywharris Something like this might be possible with NotebookLM

What’s an ai tool that allows me keep an ongoing journal of random reflections and ideas over time (years) and that I can then have a chat bot conversation that synthesizes it. one single thread that contains all of this context that I can start a thread that lasts years.