RoMa-Ω Code Released: VGGT-Ω Features Push Image Matching Past RoMa v2 on Hard Benchmarks
kwangmoo_yi · x · 2026-09-11
The authors released the code for RoMa-Ω (ECCVW 2026), a dense image matcher built on RoMa v2.
- The swap: replace RoMa v2's DINOv3 backbone with VGGT-Ω features and retrain — a small change with outsized gains.
- Results: slightly stronger than RoMa v2 on most benchmarks and considerably stronger on difficult ones, surpassing RoMa and RoMa v2 across the board; notably wins on WxBS and HardMatch, where RoMa v2 lagged.
- Positioned as a powerful but slow dense matcher; repo includes paper PDF, experiments, and source. Authors from Chalmers, ShanghaiTech, and École des Ponts.
Related event: RoMa-Ω swaps DINOv3 for feed-forward 3D features, sets matching records(3 posts)→
More from Research
- Open-source decision models sold their own base in Red Alert 2 — then learned until no human could beat them — heathercmiller · 2026-10-27
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27
- World Model Assigns Weight and Friction to Objects, Then Simulates with a Physics Engine — jonstephens85 · 2026-10-27
- Token Collapse: Study Finds Graph Soft Prompting Tokens in LLMs Fail to Encode Structure — Sapienza · 2026-10-27
- Mathathon 2.0 launches with Terence Tao's SAIR, $1.5M in compute for teams — _sathvikr · 2026-10-27
- The 'sweet lesson' is real: two kinds of ML problems need opposite strategies — mike64_t · 2026-10-27