Research map / Label-efficient and robust learning / Recognition and long-tail
HypeVPR: Exploring Hyperbolic Space for Perspective to Equirectangular Visual Place Recognition
Suhan Woo, Seongwon Lee, Jinwoo Jang, Euntai Kim
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 |
|---|---|---|---|
| embed descriptors in hyperbolic (Poincare ball) space instead of Euclidean space (ablation) | 2-level HAM hierarchy · Pitts250k-P2E | Recall@1: 9.2 -> 14.9 | Improved |
| add Euclidean triplet loss on window-level descriptors (ablation) | HypeVPR-O · Pitts250k-P2E | Recall@1: 32.0 -> 66.5 | Improved |
| add hierarchical triplet loss (ablation) | HypeVPR-O · Pitts250k-P2E | Recall@1: 50.8 -> 66.5 | Improved |
| add hyperbolic triplet loss (ablation) | HypeVPR-O · Pitts250k-P2E | Recall@1: 64.3 -> 66.5 | Improved |
| fuse top-level and lowest-level scores (HypeVPR-L) instead of lowest-level sliding window only (ablation) | HypeVPR · Pitts250k-P2E | Recall@1: 80.6 -> 81.2 | Improved |
| use hierarchical retrieval (HypeVPR-B) instead of EigenPlace exhaustive window search | EigenPlace backbone · Pitts250k-P2E | retrieval time per query (ms): 90.5 -> 17.9 | Improved |
| use only the top-level descriptor (no reranking) (ablation) | HypeVPR · Pitts250k-P2E | Recall@1: 81.2 -> 66.5 | Hurt |
| use HypeVPR-L instead of PanoVPRx16 on ConvNeXt-S | P2E baseline · Pitts250K-P2E | Recall@1: 40.3 -> 43.4 | Improved |
| use HypeVPR-L instead of PanoVPRx16 on Swin-T | P2E baseline · Pitts250K-P2E | Recall@1: 33.6 -> 32.5 | Mixed |
| HypeVPR-L versus SALAD on large-scale SF-XL | P2P baseline · SF-XL test | Recall@1: 88.6 -> 85.2 | Hurt |
Closest papers by meaning
- UniPR-3D: Towards Universal Visual Place Recognition with Visual Geometry Grounded Transformer (ECCV 2026 · similarity 0.88)
- StructVPR++: Distill Structural and Semantic Knowledge With Weighting Samples for Visual Place Recognition (TPAMI 2025 · similarity 0.88)
- EffoVPR: Effective Foundation Model Utilization for Visual Place Recognition (ICLR 2025 · similarity 0.87)
- Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition (AAAI 2025 · similarity 0.87)
- MutualVPR: A Mutual Learning Framework for Resolving Supervision Inconsistencies via Adaptive Clustering (NeurIPS 2025 · similarity 0.87)
- Towards Implicit Aggregation: Robust Image Representation for Place Recognition in the Transformer Era (NeurIPS 2025 · similarity 0.86)
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.
