XFeat Revisited: Reproducibility Study and Ablations of Lightweight Image Matcher
zhenjun_zhao · x · 2026-08-13
This paper presents a reproducibility study of XFeat, a lightweight local feature extractor and matcher. The authors re-implemented the architecture and re-evaluated the officially released checkpoint.
- Reproduction: The reproduced models closely match or outperform the original checkpoint on MegaDepth-1500 and ScanNet-1500, supporting XFeat's strong accuracy-efficiency trade-off.
- Discrepancies: The study notes several implementation differences among the original paper, supplementary material, and public code, including backbone layout, fusion blocks, and training losses.
- Ablations: Provides nuanced views on architectural claims: the parallel keypoint branch is important but less pronounced than originally claimed, while evidence for the specific skip-connection placement remains inconclusive.
More from Research
- Researchers ask: why is there no AI-first scientific journal with AI-run peer review? — anshulkundaje · 2026-09-21
- Waypoint Bio argues AI in biology has a data problem before a model problem — anshulkundaje · 2026-09-21
- New paper estimates LLMs raised US natural unemployment rate by 0.1-0.2 points — mattbeane · 2026-09-21
- JevBench open-sources a benchmark for typed decision models scoring intelligence, calibration and speed — sull · 2026-09-21
- gLM2 designs a 2,500-AA polyketide synthase making a nylon precursor, 10x titer boost — anshulkundaje · 2026-09-21
- Dev trains a world model on 50K+ real SWE-bench agent runs to guide coding agents — Decent-Ad9950 · 2026-09-21