Skip to content

Re-ID模型评估结果

预训练Resnet-18模型直接用于人车对

Dataset Rank-1 Rank-5 Rank-10 mAP mINP metric
BikePerson 0.12 0.65 1.22 0.19 0.09 0.15

VehicleID模型直接用于人车对

Dataset Rank-1 Rank-5 Rank-10 mAP mINP metric
BikePerson 0.97 2.47 3.71 0.60 0.13 0.78

Market1501 People Re-ID模型直接用于人车对

Dataset Rank-1 Rank-5 Rank-10 mAP mINP metric
BikePerson 0.93 2.55 3.80 0.62 0.13 0.77

Market1501 People Re-ID模型训练后用于人车对

Dataset Rank-1 Rank-5 Rank-10 mAP mINP metric
BikePerson 67.95 81.73 88.68 66.71 51.75 67.33

Market1501 People Re-ID模型训练后用于自行车

Dataset Rank-1 Rank-5 Rank-10 mAP mINP metric
BikePerson 56.25 71.43 79.53 52.79 34.39 54.52