BoQ: A Place is Worth a Bag of Learnable Queries
Amar Ali-bey, Brahim Chaib-Draa, Philippe Giguère
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/amaralibey/bag-of-queriesOfficialIn paperpytorch★ 173
Abstract
In visual place recognition, accurately identifying and matching images of locations under varying environmental conditions and viewpoints remains a significant challenge. In this paper, we introduce a new technique, called Bag-of-Queries (BoQ), which learns a set of global queries designed to capture universal place-specific attributes. Unlike existing methods that employ self-attention and generate the queries directly from the input features, BoQ employs distinct learnable global queries, which probe the input features via cross-attention, ensuring consistent information aggregation. In addition, our technique provides an interpretable attention mechanism and integrates with both CNN and Vision Transformer backbones. The performance of BoQ is demonstrated through extensive experiments on 14 large-scale benchmarks. It consistently outperforms current state-of-the-art techniques including NetVLAD, MixVPR and EigenPlaces. Moreover, as a global retrieval technique (one-stage), BoQ surpasses two-stage retrieval methods, such as Patch-NetVLAD, TransVPR and R2Former, all while being orders of magnitude faster and more efficient. The code and model weights are publicly available at https://github.com/amaralibey/Bag-of-Queries.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| AmsterTime | BoQ (ResNet-50) | Recall@1 | 52.2 | — | Unverified |
| AmsterTime | BoQ | Recall@1 | 63 | — | Unverified |
| Eynsham | BoQ | Recall@1 | 92.2 | — | Unverified |
| Eynsham | BoQ (ResNet-50) | Recall@1 | 91.3 | — | Unverified |
| Mapillary test | BoQ | Recall@1 | 79 | — | Unverified |
| Mapillary val | BoQ (ResNet-50) | Recall@1 | 91.2 | — | Unverified |
| Mapillary val | BoQ | Recall@1 | 93.8 | — | Unverified |
| Nordland | BoQ (ResNet-50) | Recall@1 | 83.1 | — | Unverified |
| Nordland | BoQ | Recall@1 | 90.6 | — | Unverified |
| Nordland* (2760 queries) | BoQ | Recall@1 | 81.3 | — | Unverified |
| Pittsburgh-250k-test | BoQ (ResNet-50) | Recall@1 | 95 | — | Unverified |
| Pittsburgh-250k-test | BoQ | Recall@1 | 96.6 | — | Unverified |
| Pittsburgh-30k-test | BoQ | Recall@1 | 93.7 | — | Unverified |
| Pittsburgh-30k-test | BoQ (ResNet-50) | Recall@1 | 92.4 | — | Unverified |
| San Francisco Landmark Dataset | BoQ | Recall@1 | 93.6 | — | Unverified |
| SPED | BoQ (ResNet-50) | Recall@1 | 86.5 | — | Unverified |
| SPED | BoQ | Recall@1 | 92.5 | — | Unverified |
| St Lucia | BoQ | Recall@5 | 100 | — | Unverified |
| St Lucia | BoQ (DINOv2) | Recall@1 | 100 | — | Unverified |
| SVOX Night | BoQ (ResNet-50) | Recall@1 | 87.1 | — | Unverified |
| SVOX-Overcast | BoQ (ResNet-50) | Recall@1 | 97.8 | — | Unverified |
| SVOX-Rain | BoQ (ResNet-50) | Recall@1 | 96.2 | — | Unverified |
| SVOX-Snow | BoQ (ResNet-50) | Recall@1 | 98.7 | — | Unverified |
| SVOX Sun | BoQ (ResNet-50) | Recall@1 | 95.9 | — | Unverified |
| Tokyo247 | BoQ | Recall@1 | 98.1 | — | Unverified |