arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2403.04765cs.CV
收藏

Efficient LoFTR:具备类稀疏匹配速度的半稠密局部特征匹配

Efficient LoFTR: Semi-Dense Local Feature Matching with Sparse-Like Speed

Zhejiang University(浙江大学)

Yifan Wang, Xingyi He, Sida Peng, Dongli Tan, Xiaowei Zhou

AI总结:

针对无检测器匹配器LoFTR效率低的问题,提出带自适应token选择的聚合注意力机制和两阶段相关层,在提升半稠密特征匹配精度的同时实现类稀疏匹配的速度,适配延迟敏感应用。

摘要:

我们提出了一种可高效生成图像间半稠密匹配点的新方法。此前的无检测器匹配器LoFTR在应对大视角变化和纹理匮乏场景时展现出了出色的匹配能力,但存在效率低下的问题。我们重新审视了其设计选择,并从效率和精度两方面得出了多项改进方案。一个关键发现是,由于局部信息存在共享性,在整幅特征图上执行transformer运算存在冗余,因此我们提出一种带自适应token选择的聚合注意力机制以提升效率。此外,我们发现LoFTR的精细相关模块中存在空间方差,这对匹配精度不利,为此提出了一种新颖的两阶段相关层,以实现精确的亚像素对应关系,提升精度。我们的效率优化模型比LoFTR快约2.5倍,甚至能超越当前最优的高效稀疏匹配流水线SuperPoint + LightGlue。此外,大量实验表明,与同类竞争力的半稠密匹配器相比,我们的方法能实现更高的精度,同时具备显著的效率优势。这为图像检索、三维重建等大规模或对延迟敏感的应用开辟了令人振奋的前景。项目主页:https://zju3dv.github.io/efficientloftr。

英文摘要:

We present a novel method for efficiently producing semi-dense matches across images. Previous detector-free matcher LoFTR has shown remarkable matching capability in handling large-viewpoint change and texture-poor scenarios but suffers from low efficiency. We revisit its design choices and derive multiple improvements for both efficiency and accuracy. One key observation is that performing the transformer over the entire feature map is redundant due to shared local information, therefore we propose an aggregated attention mechanism with adaptive token selection for efficiency. Furthermore, we find spatial variance exists in LoFTR's fine correlation module, which is adverse to matching accuracy. A novel two-stage correlation layer is proposed to achieve accurate subpixel correspondences for accuracy improvement. Our efficiency optimized model is $\sim 2.5\times$ faster than LoFTR which can even surpass state-of-the-art efficient sparse matching pipeline SuperPoint + LightGlue. Moreover, extensive experiments show that our method can achieve higher accuracy compared with competitive semi-dense matchers, with considerable efficiency benefits. This opens up exciting prospects for large-scale or latency-sensitive applications such as image retrieval and 3D reconstruction. Project page: https://zju3dv.github.io/efficientloftr.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑