期刊
INTERNATIONAL JOURNAL OF COMPUTER VISION
卷 129, 期 2, 页码 517-547出版社
SPRINGER
DOI: 10.1007/s11263-020-01385-0
关键词
Benchmark; Dataset; Stereo; Structure from motion; Local features; 3D reconstruction
资金
- Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant Deep Visual Geometry Machines [RGPIN-2018-03788]
- Google's Visual Positioning Service
- OP VVV [CZ.02.1.01/0.0/0.0/16 019/0000765]
- CTU [SGS17/185/OHK3/3T/13]
- Austrian Ministry for Transport, Innovation and Technology
- Federal Ministry for Digital and Economic Affairs
- Province of Upper Austria
- Swiss National Science Foundation
This study introduces a comprehensive benchmark for local features and robust estimation algorithms, focusing on the accuracy of reconstructed camera poses as the primary metric. The experiments show that classical solutions, when properly configured, may still outperform the perceived state of the art.
We introduce a comprehensive benchmark for local features and robust estimation algorithms, focusing on the downstream task-the accuracy of the reconstructed camera pose-as our primary metric. Our pipeline's modular structure allows easy integration, configuration, and combination of different methods and heuristics. This is demonstrated by embedding dozens of popular algorithms and evaluating them, from seminal works to the cutting edge of machine learning research. We show that with proper settings, classical solutions may still outperform the perceived state of the art. Besides establishing the actual state of the art, the conducted experiments reveal unexpected properties of structure from motion pipelines that can help improve their performance, for both algorithmic and learned methods. Data and code are online (https://github.com/ubcvision/image-matching-benchmark), providing an easy-to-use and flexible framework for the benchmarking of local features and robust estimation methods, both alongside and against top-performing methods. This work provides a basis for the Image Matching Challenge (https://image-matching-challenge.github.io).
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据