多方引导前景增强的行人重识别方法
|
|
刘俊婧,郑宛露,郭子强,王少荣
|
Person re-identification method based on multi-part guided foreground enhancement
|
|
Junjing LIU,Wanlu ZHENG,Ziqiang GUO,Shaorong WANG
|
|
| 表 9 提出方法与现有行人重识别方法的性能对比 |
| Tab.9 Performance comparison of proposed method with existing person re-identification method % |
|
| 方法 | Market1501 | | DukeMTMC | | Rank-1 | mAP | Rank-1 | ${\mathrm{mAP}} $ | | TriNet[21] | 84.9 | 69.1 | | 72.4 | 53.5 | | ARP[22] | 87.0 | 66.9 | 73.9 | 55.5 | | JLML[23] | 85.1 | 65.5 | — | — | | AlignedReID[24] | 92.6 | 82.1 | 81.2 | 67.4 | | HA-CNN[25] | 91.2 | 75.7 | 80.5 | 63.8 | | PCB[26] | 93.8 | 81.6 | 83.3 | 69.2 | | Mancs[27] | 93.1 | 82.3 | 84.9 | 71.8 | | OSNet[28] | 94.8 | 84.9 | 87.6 | 73.5 | | Auto-ReID[29] | 94.5 | 85.1 | 88.5 | 75.1 | | HOReid[30] | 94.2 | 84.9 | 86.9 | 75.6 | | CBDB-Net[31] | 94.4 | 85.0 | 87.7 | 74.3 | | DRL-Net[32] | 94.7 | 86.9 | 88.1 | 76.6 | | MGFE(本文方法) | 95.0 | 87.5 | 88.3 | 77.9 |
|
|
|