GUSH3R: Single Forward Pass for Dynamic Human-Scene 3D Reconstruction
rsasaki0109 · x · 2026-08-04
Researchers have introduced GUSH3R, a feed-forward framework for online dynamic human-scene reconstruction. It tackles the challenging problem of reconstructing dynamic environments from monocular videos.
Previous feed-forward methods often predicted non-photorealistic representations (like point clouds and meshes) or failed to handle non-rigid objects, particularly dynamic humans. Utilizing 3D Gaussian Splatting (3DGS) primitives, GUSH3R jointly models dynamic humans and static scenes in a single forward pass, enabling photorealistic rendering of complex non-rigid motions.
More from Multimodal
- Seedance Model Family Lands Exclusively on Dreamina for Accessible AI Video Creation — AIwithGhotai · 2026-08-04
- MiniMax Video Workflow: How to Maintain Character Consistency in Longer Clips? — Like_Zorro · 2026-08-04
- RTX 5090 Benchmark: Generates 1-Megapixel 21:9 Video in 4 Minutes — AdmirablePainting368 · 2026-08-04
- Vidu's One-Click MV Upgrade: Introduces Lip-Sync and Dance Choreography Pipelines — 生数科技 · 2026-08-04
- AI Recreates Viral Illusion: 100+ Indian Folk Dancers Form a Giant Elephant — sivakumaranvlsi · 2026-08-04
- Minimax-H3 Multi-Precision Quantized Version Hits HF Trending, Supports ComfyUI — Abiray · 2026-08-04