RWTH Aachen's ARROW unifies 3D reconstruction and point tracking from any RGB inputs, sets new SOTA
kwangmoo_yi · x · 2026-10-09
RWTH Aachen researchers released ARROW, a feed-forward model unifying 3D reconstruction and 3D point tracking from arbitrary image sets — photo collections, moving-camera videos, or synchronized streams — without temporal ordering, thanks to a novel order-invariant querying approach. Training on diverse input mixtures improves performance and enables multi-view tracking generalization. ARROW sets new SOTA in 3D tracking on WorldTrack and TAPVid-3D and beats dedicated multi-view trackers on an adapted RGB-only MVTracker benchmark while staying competitive on 3D reconstruction.
More from Multimodal
- AI image of the day: GPT-6 Image 2.5 recreates 1940s Chinese village newsreel — DeryaTR_ · 2026-10-09
- Vigglorious Studio: chunked reference frames and guide keyframes fix character drift in long AI videos — Tablaski · 2026-10-09
- Star Wars AI songs strike again: Qui-Gon gets his own 'Gin and Juice' anthem — aronchick · 2026-10-09
- MiniMax H3 zombie short film goes viral on Reddit — gabxav · 2026-10-09
- Seedance 2.5 AI video praised as a "masterpiece" in viral share — BLUECOW009 · 2026-10-09
- HeyGen Video Tops OpenRouter's Video Rankings, Hands Out Codes for 100 Free Clips Each — aziz4ai · 2026-10-09