Research map / Architectures and efficiency / Attention and transformers
R2Former: Unified Retrieval and Reranking Transformer for Place Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use R2Former instead of TransVPR | VPR retrieval and reranking · MSLS Val | R@1: 86.8 -> 89.7 | Helped |
| Use R2Former instead of TransVPR | VPR retrieval and reranking · MSLS Challenge | R@1: 63.9 -> 73.0 | Helped |
| Use R2Former instead of TransVPR | VPR retrieval and reranking · Pitts30k | R@1: 89.0 -> 91.1 | Helped |
| Use R2Former instead of TransVPR | VPR retrieval and reranking · Tokyo 24/7 | R@1: 79.0 -> 88.6 | Helped |
| Add the learned R2Former reranker | R2Former global retrieval on MSLS Val · MSLS Val | R@1: 79.3 -> 89.7 | Helped |
| Add global hard-negative mining and joint finetuning after reranker training (ablation) | R2Former training · MSLS Val | R@1: 86.6 -> 89.7 | Helped |
| Train retrieval and reranking end-to-end from the beginning (ablation) | R2Former training · MSLS Val | R@1: 79.3 -> 87.3 | Helped |
| Add positional embedding (ablation) | R2Former reranking module · MSLS Val | R@1: 86.4 -> 86.6 | No clear effect |
| Add Transformer Block-1 (ablation) | R2Former reranking module · MSLS Val | R@1: 85.5 -> 86.6 | Helped |
| Add Transformer Block-2 (ablation) | R2Former reranking module · MSLS Val | R@1: 81.9 -> 86.6 | Helped |
| Use Top-5 feature pairs instead of Top-1 (ablation) | R2Former reranking module · MSLS Val | R@1: 86.6 -> 86.6 | No clear effect |
| Use attention-based feature selection (ablation) | R2Former reranking module · MSLS Val | R@1: 83.6 -> 86.6 | Helped |
Closest papers by meaning
- Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition (AAAI 2025 · similarity 0.92)
- Towards Implicit Aggregation: Robust Image Representation for Place Recognition in the Transformer Era (NeurIPS 2025 · similarity 0.90)
- EffoVPR: Effective Foundation Model Utilization for Visual Place Recognition (ICLR 2025 · similarity 0.88)
- BoQ: A Place is Worth a Bag of Learnable Queries (CVPR 2024 · similarity 0.88)
- Optimal Transport Aggregation for Visual Place Recognition (CVPR 2024 · similarity 0.88)
- MutualVPR: A Mutual Learning Framework for Resolving Supervision Inconsistencies via Adaptive Clustering (NeurIPS 2025 · similarity 0.87)
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.