Cross-modal alignment without paired data.

ARROW unifies 3D reconstruction and point tracking from arbitrary RGB observations through order-invariant querying and training across viewpoints and capture times.

A plug-in that densifies the geometry tokens of pretrained feed-forward 3DGS models, so a frozen backbone reconstructs as if it had seen many more views than it was given.

Reading the visual semantics that emerge inside multimodal diffusion transformers, then strengthening them to improve generation.

Can a coding agent watch a video of a physical event and write the program that rebuilds it? 200 tasks, 100 real videos and 100 simulated.

Standard 3D Gaussian Splatting (3DGS) learns geometry and appearance jointly from RGB supervision, making it difficult to isolate how luminance and chroma contribute to the learned representation. We ...

Frontier coding agents can now write and execute code that authors 3D environments, but whether they reliably understand 3D structure and precisely control scene state remains unclear. The generated 3...

ProDyGS is a method for dynamic Gaussian splatting from a single static monocular camera, achieving state-of-the-art performance in novel view synthesis.

We present a new inference-time sampler for diffusion models that gives a pretrained model a capability it was never trained for: constructing the atlas of the population it synthesizes. The sampler c...

ICP-based 3D Gaussian Splatting (3DGS) SLAM tracks in real time by registering incoming frames against map Gaussians, using each primitive's covariance for both rendering and registration. These two u...

official repo for RIGOR: Rig-Informed Geometry for Omnidirectional Reconstruction - TangentH/RIGOR

We present DRS-VPT, a feed-forward transformer architecture for foundational image-to-scan registration. Given query images and a reference 3D point cloud, the model predicts the scan pose and point m...

[ECCVW 2026] RoMa-Ω: What Feed-Forward 3D Models Know About Image Matching - davnords/RoMa-Omega

Depth, surface normals, and albedo from a single image

Contribute to naver/blast3r development by creating an account on GitHub.

Generating complete 3D scenes from sparse, unconstrained views is a fundamental challenge in 3D vision which requires reasoning beyond observed content while remaining computationally tractable. Exist...

SolarWM — Open Data and Scalable Training for Long-Horizon Video World Models

[ECCV 2026] Efficient and High-Quality Depth Estimation via Pixel-Space Diffusion with Linear Attention - VISION-SJTU/Lapis

VATIX: A Scaling Law Analysis of Video Diffusion Models for driving video generation.

While state-of-the-art generative models produce high-fidelity 3D meshes, these outputs lack the physical properties required for interactive simulation, gaming, or robotics. We introduce Gen2Physics,...

One video model repairs rendering artifacts from 3DGS, NeRF, mesh, or sparse point clouds. ECCV 2026, Carnegie Mellon University.

Official implementation of MVAP-G [ECCV'26 Oral]. Contribute to qsong2001/mvap-g development by creating an account on GitHub.

Equilibrium Forcing: Adaptive Video Generation Without Noise Conditioning

Contribute to ThereWillComeSoftRains/RevisitingCFGMethods development by creating an account on GitHub.