Deep Camera-Aware Metric Learning for Person Reidentification

Table 2

Comparison results (%) on Market1501 dataset where the bold font and underline denotes the best method; bold font denotes the 2nd best method, and underline denotes the 3rd best method. These methods explore part-based features for a person re-id. Our proposed model achieves the 4th best results on all the 2 evaluation metrics.

Method	Reference	Rank-1	mAP
Ours	—	92.7	78.7
MGN	ACMMM [33]	95.7	88.2
HPM	AAAI [16]	94.2	82.7
PCB+RPP	ECCV [15]	93.8	81.6
PCB	ECCV [15]	92.3	77.4
GLAD	ACMMM [34]	89.9	73.9
Partloss	TIP [35]	88.2	59.3
PDC	ICCV [11]	84.4	63.4
Multiloss	IJCAI [36]	83.9	64.4
PAR	ICCV [9]	81.0	63.4
Hydra	ICCV [12]	76.9	—
MRegion	AVSS [37]	66.4	41.2