The Image Matching Challenge (IMC) Phototourism benchmark (IMC-PT) is part of the Image Matching Challenge / Image Matching Workshop (IMW) benchmark suite used to evaluate local features, matching methods and robust geometry estimation for camera pose estimation and multi-view reconstruction. The benchmark provides image collections (Phototourism scenes) with Colmap-derived (pseudo-)ground-truth camera poses, densified depth maps and co-visibility information; it is organized into tracks for stereo (image pairs) and multi-view reconstruction and supports restricted/unrestricted keypoint settings. Typical evaluation measures include camera-pose accuracy summarized as AUC of angular/positional thresholds (reported commonly as AUC@3°, AUC@5°, AUC@10°), plus runtime. The IMC site and the UBC-hosted pages provide downloadable training/validation sets (photo-tourism scenes), leaderboards, and an open-source evaluation pipeline (github: ubc-vision/image-matching-benchmark). The benchmark has been used in the IMW/CVPR workshop challenges (2020/2021 etc.) and is the reference benchmark for many recent local-feature and image-matching papers.
No results indexed yet — be the first to submit a score.
Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.