Research Article

Circle-Based Ratio Loss for Person Reidentification

Table 3

The comparisons of our proposed method with LMCL and ArcFace on Market-1501, DukeMTMC-reID, and CUHK03 datasets.

MethodsmMarket-1501DukeMTMC-reIDCUHK03 labeledCUHK03 detected
mAPRank-1mAPRank-1mAPRank-1mAPRank-1

LMCL [21]0.0182.4992.7071.8784.2963.1263.7161.4263.36
0.182.9393.1172.3284.9263.5566.5062.5463.64
0.382.0592.4971.4383.9363.3264.6460.7162.00
0.581.7093.0270.6884.0762.1564.2158.4960.43
1.081.1592.2570.5184.1161.8663.5058.4659.79

ArcFace [5]0.0182.3192.9971.1984.0263.0264.2160.9263.00
0.182.7092.9671.2883.9364.3366.5062.5464.93
0.382.5292.9971.0984.7865.2067.5762.6563.43
0.581.2892.3470.5483.8964.6265.8661.8363.07
1.080.8691.4269.6083.0862.7664.1460.7261.50

Ours—83.1292.6471.6684.3466.2668.5763.2465.07

The bold values indicate the best results of all the methods on each metric. They are beneficial to compare between our proposed method and the other methods.