📢 We are thrilled to announce the Call for Papers for #Eurographics2026 in Aachen! 🇩🇪📷Join us in shaping the future of computer graphics - submit your cutting-edge research! Deadline: Sept. 26, 2025 Check it out: eg2026.github.io/call_for_full_…

Do not miss Dynamic Neural Point Clouds, our #EUROGRAPHICS2025 full paper that lifts monocular non-rigid view synthesis to a new level by combining fast optimisation, real-time frame rates and highly competitive rendering quality. arxiv.org/pdf/2406.10078 github.com/MoritzKappel/D…

thomas strikes back - new jit based tiny-cuda-nn release allows instantngp to render 5x faster and train 1.5x faster with less code. cool to see how far you (or thomas :)) can push kernel fusion github.com/NVlabs/instant…

For everyone interested in precise 📷camera control 📷 in transformers [e.g., video / world model etc] Stop settling for Plücker raymaps -- use camera-aware relative PE in your attention layers, like RoPE (for LLMs) but for cameras! Paper & code: liruilong.cn/prope/

I honestly do not care very much about one of my papers being accepted and rejected... This happens ~5% of the time and it gets averaged out... But for a first year student, that threw their entire soul to write the best paper they can... this is soul crushing.

Earlier today, my colleague Junbo finished the release for our work on ray-based GS from spherical panoramas. Comes with a cool dataset and an efficient method to compute exact BBs on the equirectangular image plane for ray-based evaluation. More info at junboli-cn.github.io/spags/

VGGT has been re-licensed to allow commercial usage. Enjoy the gift 😉😇 👉 huggingface.co/facebook/VGGT-…

July has been a big month for Viser! - Released v1.0.0😊 - We did some writing Some demos👇

The first bounty is complete, and we have a winner: Florian Hahlbohm, achieving a ~2.4x speedup over the baseline implementation. The results are impressive: a runtime of 20m 22s on the mipnerf360 dataset using MCMC, with 1M Gaussians and 30k iterations on an RTX 4090. This leaves

[SIGGRAPH '25] Echoes of the Coliseum: Towards 3D Live streaming of Sports Events TL;DR: Towards live streaming sport events as 3D Gaussian Splats! Contributions: 1. A scalable system design for real-time reconstruction of sports events via distributed and parallel processing.

Dr.Jit+Mitsuba just added support for fused neural networks, hash grids, and function freezing to eliminate tracing overheads. This speeds up optimization & realtime jobs and enables custom Instant NGP and neural material/radiosity/path guiding projects. What will you do with it?

Methods like NeRF and Gaussian Splats model the world as radioactive fog, rendered using alpha blending. This produces great results.. but are volumes the only way to get there?🤔 Our new SIGGRAPH'25 paper directly reconstructs surfaces without heuristics or regularizers.

Official Launch of the MrNeRF 3DGS Bounty 2: We're offering 🏆 $1600 + $500 bonus for improving initialization & training without densification for 3D Gaussian Splatting! RT & tag friends who might crush this. Details in thread 👇

Posting macro splats of bugs to SuperSplat is the new trend in graphics! 🦋 The way that spherical harmonics encode the iridescence of the butterfly's wings is spectacular. [1 / 3]

[SIGGRAPH '25] Echoes of the Coliseum: Towards 3D Live streaming of Sports Events TL;DR: "..., an end-to-end streaming framework, which takes multi-view video streams of sports events as input and reconstructs the dynamic 3D scene to allow viewers to watch from any novel view

The @PlayCanvas team is proposing an update to the SOGS format for 3D Gaussian Splatting. If you care about 3DGS and file formats, please join the discussion. We want your feedback! 👂 github.com/playcanvas/spl…

📢Code release for 3DGS-LM (#ICCV2025)📢 Gaussian-Splatting made faster by changing the underlying optimizer. It's compatible with other changes that accelerate the optimization along different axes. lukashoel.github.io/3DGS-LM/

image2world

tiny-cuda-nn JIT fusion is amazing. Don't sleep on it! Just used it to integrate a tcnn.NetworkWithInputEncoding into a 3DGS forward pass, and holy sh*t it's fast.

Really cool paper, check it out! Happy to see people building upon HTGS. An undervalued property of improved multi-view consistency is that you get the same quality with less primitives. With the improved culling and anti-aliasing techniques from this paper you can go even lower.