📢 Looking for 1 PhD student (3 years) in Vienna! We have an open position for someone to tackle MASSIVE 3D datasets and turn them into instant visuals for display, editing, and interacting! You will be supervised by Vienna's Dr. Schütz and myself (remote). Link in comment.

📢📢📢 𝐒𝐩𝐡𝐞𝐫𝐢𝐜𝐚𝐥 𝐕𝐨𝐫𝐨𝐧𝐨𝐢: directional appearance as a differentiable partition of the sphere – Project: sphericalvoronoi.github.io – Paper: arxiv.org/abs/2512.14180 – Colab: tinyurl.com/spherical-vorono… Project led by 𝐅𝐫𝐚𝐧𝐜𝐞𝐬𝐜𝐨 𝐝𝐢 𝐒𝐚𝐫𝐢𝐨 (graduating!)

I’m heading to SIGGRAPH Asia 2025 in #HongKong this 15-18 Dec! Looking forward to connecting with peeps & exploring the latest in graphics and visual AI!

Introducing Generative View Stitching (GVS), a non-autoregressive sampling method for length extrapolation of video diffusion models. GVS enables collision-free camera-guided video generation for predefined trajectories, including Oscar Reutersvärd's Impossible Staircase (1/9).

Heading to ICCV to see the latest and craziest in visual computing and machine learning! My schedules are wide open and I would love to chat about anything research, feel free to reach out!

Checkout our new SIGGRAPH Asia 2025 paper: repo-sam.inria.fr/nerphys/ed…. For more details see linkedin.com/posts/george-dr…

RT @taiyasaki: 📢📢📢 𝐓𝐫𝐢𝐚𝐧𝐠𝐥𝐞 𝐒𝐩𝐥𝐚𝐭𝐭𝐢𝐧𝐠 : 𝐃𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐭𝐢𝐚𝐛𝐥𝐞 𝐑𝐞𝐧𝐝𝐞𝐫𝐢𝐧𝐠 𝐰𝐢𝐭𝐡 𝐎𝐩𝐚𝐪𝐮𝐞 𝐓𝐫𝐢𝐚𝐧𝐠𝐥𝐞𝐬. – Project: trianglesplatting2.github.io – ArXiv: https…

Janusch, take the wheel

We are very excited to announce our work "On-the-fly Reconstruction for Large-Scale Novel View Synthesis from Unposed Images", to be presented at #SIGGRAPH2025! Learn more: repo-sam.inria.fr/nerphys/on… 1/3

StochasticSplats: Stochastic Rasterization for Sorting-Free 3D Gaussian Splatting Abstract: 3D Gaussian splatting (3DGS) is a popular radiance field method with many application-specific extensions. Most variants rely on the same core algorithm: depth-sorting of Gaussian splats followed by rasterizing in primitive order. This ensures correct alpha compositing but can cause rendering artifacts due to built-in approximations. Moreover, for a fixed representation, sorted rendering offers little control over render cost and visual fidelity. For example, counter-intuitively, rendering a lower-resolution image is not necessarily faster. In this work, we address these limitations by combining 3D Gaussian splatting with stochastic rasterization. Concretely, we leverage an unbiased Monte Carlo estimator of the volume rendering equation. This removes the need for sorting and allows for accurate 3D blending of overlapping Gaussians. The number of Monte Carlo samples further imbues 3DGS with a way to trade off computation time and quality. We implement our method using OpenGL shaders, enabling efficient rendering on modern GPU hardware. At a reasonable visual quality, our method renders more than four times faster than sorted rasterization.

Two days after the h-1b fallout, and 80% of the nazi stuff has vanished from my twitter feed. Interesting?

This project blew my mind! It looks just as good as 3DGS but it has no per-scene optimization or reconstruction. Every frame is generated *from scratch* and it nevertheless runs in real-time. quark-3d.github.io/

The original 3DGS code is an academic research project, the fact that somebody is still actively improving it 1.5 years later is 🧑‍🍳😘 (chef's kiss)

Hot dang!