Pair-VPR: Place-Aware Pre-training and Contrastive Pair Classification for Visual Place Recognition with Vision Transformers

Open paper
Benchmark Model Rank Results
visual-place-recognition-on-mapillary-testPair-VPR-p#1Recall@1: 81.7Recall@5: 90.2Recall@10: 91.3
visual-place-recognition-on-mapillary-testPair-VPR-s#4Recall@1: 79.0Recall@5: 86.9Recall@10: 88.3
visual-place-recognition-on-mapillary-valPair-VPR-p#1Recall@1: 95.4Recall@5: 97.3Recall@10: 97.7
visual-place-recognition-on-mapillary-valPair-VPR-s#3Recall@1: 93.7Recall@5: 97.2Recall@10: 97.3
visual-place-recognition-on-pittsburgh-30kPair-VPR-p#1Recall@1: 95.4Recall@5: 97.5Recall@10: 98.0
visual-place-recognition-on-pittsburgh-30kPair-VPR-s#2Recall@1: 94.7Recall@5: 97.2Recall@10: 97.8
visual-place-recognition-on-tokyo247Pair-VPR-p#1Recall@1: 100Recall@5: 100Recall@10: 100
visual-place-recognition-on-tokyo247Pair-VPR-s#4Recall@1: 98.1Recall@5: 98.4Recall@10: 98.7

Rank counts only results with a code link.