Research map / 3D vision / Depth and stereo
MeshVPR: Citywide Visual Place Recognition Using 3D Meshes
Gabriele Berton, Lorenz Junglas, Riccardo Zaccone, Thomas Pollok, Barbara Caputo, Carlo Masone
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 |
|---|---|---|---|
| use a synthetic mesh-rendered database instead of a real-image database | SALAD · San Francisco | Recall@1: 96.3 -> 76.9 | Hurt |
| add MeshVPR features alignment to SALAD on synthetic database | SALAD · Synt-Berlin | Recall@1: 69.8 -> 82.0 | Improved |
| add MeshVPR features alignment to EigenPlaces on synthetic database | EigenPlaces · Synt-Melbourne | Recall@1: 26.7 -> 47.6 | Improved |
| add MeshVPR features alignment to Conv-AP on synthetic database | Conv-AP · Synt-Paris | Recall@1: 34.3 -> 35.4 | No clear effect |
| add MeshVPR features alignment to EigenPlaces on synthetic database | EigenPlaces · Synt-Paris | Recall@1: 44.8 -> 48.5 | Improved |
| use high-quality mesh instead of low-quality mesh for alignment and test (ablation) | SALAD + MeshVPR · San Francisco | Recall@1: 76.1 -> 88.8 | Improved |
| train alignment on HQ mesh and test on LQ database instead of training on LQ (ablation) | MixVPR + MeshVPR · San Francisco LQ DB | Recall@1: 59.7 -> 53.0 | Hurt |
| use features alignment with two models instead of metric-learning fine-tuning on mixed real and synthetic data (ablation) | EigenPlaces · Synt-Berlin | Recall@1: 40.4 -> 67.1 | Improved |
| increase alignment training set from 100k to 9.2M image pairs (ablation) | MeshVPR · Synt-Berlin | Recall@1: 81.2 -> 82.0 | No clear effect |
| increase alignment training set from 10k to 100k image pairs (ablation) | MeshVPR · Synt-Paris | Recall@1: 59.3 -> 61.2 | Improved |
| add MeshVPR to SALAD on synthetic San Francisco database | SALAD · San Francisco | Recall@1: 76.9 -> 88.8 | Improved |
Closest papers by meaning
- Deep Homography Estimation for Visual Place Recognition (AAAI 2024 · similarity 0.88)
- Towards Seamless Adaptation of Pre-trained Models for Visual Place Recognition (ICLR 2024 · similarity 0.88)
- Towards Implicit Aggregation: Robust Image Representation for Place Recognition in the Transformer Era (NeurIPS 2025 · similarity 0.87)
- SuperVLAD: Compact and Robust Image Descriptors for Visual Place Recognition (NeurIPS 2024 · similarity 0.87)
- GPVK-VL: Geometry-Preserving Virtual Keyframes for Visual Localization under Large Viewpoint Changes (CVPR 2025 · similarity 0.87)
- Optimal Transport Aggregation for Visual Place Recognition (CVPR 2024 · 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.
