Sample4Geo: Hard Negative Sampling For Cross-View Geo-Localisation

Benchmark Model Rank Results
cross-view-geo-localisation-on-spagbolSample4Geo#2Top-1: 50.80
drone-view-target-localization-on-university-1Sample4Geo#2AP: 93.81Recall@1: 92.65
image-based-localization-on-cvactSample4Geo#2Recall@1: 90.81Recall@5: 96.74Recall@10: 97.48Recall@1 (%): 98.77
image-based-localization-on-cvusa-1Sample4Geo#2Recall@1: 98.68Recall@5: 99.68Recall@10: 99.78Recall@top1%: 99.87
image-based-localization-on-vigor-cross-areaSample4Geo#2Recall@1: 61.70Recall@5: 83.50Recall@10: 88.00Recall@1%: 98.17
image-based-localization-on-vigor-same-areaSample4Geo#2Recall@1: 77.86Recall@5: 95.66Recall@10: 97.21Recall@1%: 99.61
visual-place-recognition-on-cv-citiesSample4Geo#2Recall@1: 74.49Recall@5: 84.07