Pumpire: Unified Benchmark for Metric Distance Estimation
Pumpire benchmarks point-pair metric distance accuracy for image and video 3D foundation models.
Researchers introduce Pumpire, a benchmark for metric point-pair distance estimation by image- and video-level 3D foundation models, with or without depth priors. The pumpire-6k set contains 100 real-world scenes, each annotated with physically measured point-pair distances and 64 frames, for 6,400 frames total. Unlike separate depth or geometric-similarity scores, the protocol measures point-to-point distance error directly. Experiments cover 29 baseline configurations of representative 3D foundation models.
- Pumpire scores point-to-point metric distance, not separate depth or similarity metrics.
- pumpire-6k has 100 measured scenes and 6,400 frames.
- Evaluation covers image- and video-level 3D models with or without depth priors.
- Authors compare 29 baseline configurations of 3D foundation models.
From the original article
1 min read at huggingface.co
We present Pumpire, a unified benchmark for evaluating metric point-pair distance estimation capability of both image- and video-level 3D foundation models, with or without depth priors.
By Siyu Chen, Zehan Wang, Jiayang Xu · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.