📢 New paper CVPR 25!
Can meshes capture fuzzy geometry? Volumetric Surfaces uses adaptive textured shells to model hair, fur without the splatting / volume overhead. It’s fast, looks great, and runs in real time even on budget phones.
🔗 autonomousvision.github.io/volsurfs/
📄 arxiv.org/pdf/2409.02482

AWESOME course (with videos) 🚨 #MachineLearning for Inverse #Graphics with @vincentsitzmann.bsky.social 💪

Includes a lecture on Gaussian Splatting Course: www.scenerepresentations.org/courses/2023...

I will be presenting Gaussian Garments next week at #3DV2025

Thursday, March 27, evening poster session, poster #17

Don't miss!

🌐 Project: eth-ait.github.io/Gaussian-Gar...
📜 Paper: arxiv.org/abs/2409.08189
🧑‍💻 Github: github.com/eth-ait/Gaus...

It was an amazing collaboration between @ethzurich.bsky.social and MPI-IS

Thanks to: Boxiang Rong, Wenbo Wang, michael-j-black.bsky.social, Bernhard Thomaszewski, Christina Tsalicoglou, Otmar Hilliges.

The reconstructed garments can then be combined into complex multi-garment outfits.

🔎 Here we drape five garments on top of each other and simulate them together with ContourCraft, our previous work.

Then, it optimizes the garment’s appearance. The appearance is represented by an albedo Gaussian texture and an appearance model that predicts lighting effects.

🔎 Note that the shadows and specular effects are predicted by the appearance model and are not baked into the texture.

First, the Gaussian Garments pipeline reconstructs the geometry of a garment and registers it to multi-view videos.

The registration procedure uses 3D Gaussian splatting and can successfully register highly dynamic sequences.

🎉🎉🎉 Happy to announce that the code for our paper Gaussian Garments is now public!

Link: github.com/eth-ait/Gaus...

Gaussian Garments uses a combination of 3D meshes and Gaussian splatting to reconstruct photorealistic simulation-ready digital garments from multi-view videos. 🧵

Graph Transformers (GTs) can handle long-range dependencies and resolve information bottlenecks, but they’re computationally expensive. Our new model, Spexphormer, helps scale them to much larger graphs – check it out at NeurIPS next week, or the preview here!
[1/13]
#NeurIPS2024

My deep learning course at the University of Geneva is available on-line. 1000+ slides, ~20h of screen-casts. Full of examples in PyTorch.

fleuret.org/dlc/

And my "Little Book of Deep Learning" is available as a phone-formatted pdf (nearing 700k downloads!)

fleuret.org/lbdl/

For those who missed this post on the-network-that-is-not-to-be-named, I made public my "secrets" for writing a good CVPR paper (or any scientific paper). I've compiled these tips of many years. It's long but hopefully it helps people write better papers. perceiving-systems.blog/en/post/writ...