Research map / Label-efficient and robust learning / Recognition and long-tail
Learning and Calibrating Per-Location Classifiers for Visual Place Recognition
Petr Gronát, Josef Šivic, Guillaume Obozinski, Tomáš Pajdla
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add calibration (w-norm) to learnt per-location SVMs (ablation) | BOW per-location SVM · Pittsburgh 25k (BOW) | recall@1: 6.4 -> 31.8 | Improved |
| add p-val calibration to learnt per-location SVMs (ablation) | BOW per-location SVM · Pittsburgh 25k (BOW) | recall@1: 6.4 -> 33.0 | Improved |
| learn per-location classifiers with p-val calibration | BOW · Pittsburgh 25k (BOW) | recall@1: 28.7 -> 33.0 | Improved |
| learn per-location classifiers with w-norm calibration | BOW · Pittsburgh 25k (BOW) | recall@1: 28.7 -> 31.8 | Improved |
| use w-norm instead of confuser suppression | BOW · Pittsburgh 25k (BOW) | recall@1: 29.6 -> 31.8 | Improved |
| use p-val instead of w-norm calibration for larger K | BOW · Pittsburgh 25k (BOW) | recall@20: 66.4 vs 69.4 | Mixed |
| learn w-norm per-location classifiers | FV128 · Pittsburgh 25k (FV) | recall@1: 33.6 -> 38.3 | Improved |
| learn w-norm per-location classifiers | FV2048 · Pittsburgh 25k (FV) | recall@1: 46.9 -> 50.2 | Improved |
| learn w-norm per-location classifiers | FV128 · Pittsburgh 55k (FV) | recall@1: 10.9 -> 13.5 | Improved |
| learn w-norm per-location classifiers | FV2048 · 24/7 Tokyo (FV) | recall@1: 37.4 -> 42.9 | Improved |
| use p-val calibration instead of w-norm (ablation) | FV128 · Pittsburgh 25k (FV) | recall@1: 33.6 -> 25.3 | Hurt |
| use LDA classifiers instead of hinge-loss SVM with hard negatives (ablation) | FV128 · Pittsburgh 25k (FV) | recall@1: 38.3 -> 31.9 | Hurt |
Closest papers by meaning
- Optimal Transport Aggregation for Visual Place Recognition (CVPR 2024 · similarity 0.84)
- MutualVPR: A Mutual Learning Framework for Resolving Supervision Inconsistencies via Adaptive Clustering (NeurIPS 2025 · similarity 0.84)
- Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition (AAAI 2025 · similarity 0.84)
- BoQ: A Place is Worth a Bag of Learnable Queries (CVPR 2024 · similarity 0.84)
- Towards Implicit Aggregation: Robust Image Representation for Place Recognition in the Transformer Era (NeurIPS 2025 · similarity 0.83)
- Towards Seamless Adaptation of Pre-trained Models for Visual Place Recognition (ICLR 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
