NVIDIA’s AI watched 150,000 videos… and learned to relight scenes incredibly well! No game engine. No 3D software. And it has an amazing cat demo. 🐱💡 Hold on to your papers! Full video: youtube.com/watch?v=yRk6vG…

Thought I'd share this WebGL viewer that uses a combination of ray tracing and depth testing to render 3D (or 2D) Gaussians. github.com/fhahlbohm/dept… Runs smoothly (>120 Hz) on an M1 MacBook Pro at 1080p. Quality is decent. Gaussians truncated at 2σ. No higher degree SH support.

Sadly, I am no longer a professor at ETH (@eth_en) due to very severe #longCovid and #MECFS. ethrat.ch/de/ernennungen….

Time to throw away the Plücker raymaps - a really elegant formulation of a camera-aware RoPE like embedding for multiview ViTs! Great work by Ruilong Li Junchen Liu and the team!

At ICLR 2024, we proposed GTA to show that relative positional encodings are better in 3D Vision Transformers. GTA had the shortcoming that it only used extrinsics. Li et. al have now fixed this, incorporating intrinsics as well. Essential advance for 3D Vision Transformers!

Had a great time chatting with Sophia and the BuzzRobot community about our recent work, DiffusionRenderer, and the exciting research my team is doing at NVIDIA AI! DiffusionRenderer project page: research.nvidia.com/labs/toronto-a…

Super cool!

I strongly agree with this concern — the policy prohibiting external links, images, and videos in the rebuttal puts vision papers at a clear disadvantage. Reviewers will tend to simply reject submissions, noting that the requested visualizations were not provided. NeurIPS Conference

GSCache: Real-Time Radiance Caching for Volume Path Tracing using 3D Gaussian Splatting Contributions: • We introduce a novel radiance cache optimized for volume rendering that caches path-space radiance using multiple levels of Gaussian splats. • The cache works in real time

Glad to see gsplat powering real-time physics simulation together with NVIDIA Warp! Learn more 👉 nvda.ws/4m4LGnN

3DGRT is now available also with Vulkan API as part of the Vulkan Gaussian Splatting: github.com/nvpro-samples/…

Very handy and well-engineered toolbox run with speed-of-light. Led by Jiahui Huang at NVIDIA AI. Project page: research.nvidia.com/labs/toronto-a…

[1/N] 🎥 We've made available a powerful spatial AI tool named ViPE: Video Pose Engine, to recover camera motion, intrinsics, and dense metric depth from casual videos! Running at 3–5 FPS, ViPE handles cinematic shots, dashcams, and even 360° panoramas. 🔗 research.nvidia.com/labs/toronto-a…

So proud to see the DLSS teams’ work come to life in Blender — such a labor of love from so many people!

📢New efficient Hybrid-SLM from NVIDIA-Nemotron-Nano-v2-9B: ❗️6x faster than Qwen3-8B because of Hybrid (Mamba2+Attention) design. We tried something new: pretrain & align a 12B reasoning model → compress to 9B. First real stab at reasoning-model compression. Key takeaways

Great to see 3DGUT now integrated into LichtFeld Studio — the fastest 3DGS/GUT codebase out there. Open source keeps pushing boundaries. Huge congrats to everyone involved! 🎉