全局学习扩展的可见光-红外行人重识别

郭子强,肖璇,陶浩然,王少荣

Global learning-expanded visible-infrared person re-identification

Ziqiang GUO,Xuan XIAO,Haoran TAO,Shaorong WANG

表 1 SYSU-MMO1数据集上GLE与先进方法的性能比较

Tab.1 Performance comparison of GLE and state-of-the-art methods on SYSU-MM01 dataset

方法 全搜索 室内搜索 R-1/% R-10/% R-20/% mAP/% R-1/% R-10/% R-20/% mAP/% DART[30] 68.7 96.4 99.0 66.3 72.5 97.8 99.5 78.2 CAJ[31] 69.9 95.7 98.5 66.9 76.3 97.9 99.5 80.4 MPANet[32] 70.6 96.2 98.8 68.2 76.7 98.2 99.6 81.0 MMN[33] 70.6 96.2 99.0 66.9 76.2 97.2 99.3 79.6 DCLNet[34] 70.8 — — 65.3 73.5 — — 76.8 MAUM[4] 71.7 — — 68.8 77.0 — — 81.9 DEEN[16] 74.7 97.6 99.2 71.8 80.3 99.0 99.8 83.3 HOS-Net[35] 75.6 — — 74.2 84.2 — — 86.7 SAAI[36] 75.9 — — 77.0 83.2 — — 88.0 MUN[37] 76.2 97.8 — 73.8 79.4 98.1 — 82.1 MSCLNet[38] 77.0 97.6 99.2 71.6 78.5 99.3 99.9 81.2 PartMix[17] 77.8 — — 74.6 81.5 — — 84.4 IDKL[18] 81.4 97.4 98.9 79.9 87.1 98.3 99.3 89.4 GLE 78.9 98.4 99.6 76.1 86.0 99.3 99.7 88.1