notesum.ai

Published at November 27

When Large Vision-Language Models Meet Person Re-Identification

cs.CV

Released Date: November 27, 2024

Authors: Qizao Wang¹, Bin Li¹, Xiangyang Xue¹

Aff.: ¹Fudan University

Arxiv: http://arxiv.org/abs/2411.18111v1

Refer to caption

Backbone	Methods	DukeMTMC-reID		Market-1501		CUHK03
Backbone	Methods	mAP	Rank-1	mAP	Rank-1	mAP	Rank-1
CNN	MGN [24]	78.4	88.7	86.9	95.7	67.4	68.0
	DG-Net [36]	74.8	86.6	86.0	94.8	-	-
	SAN [8]	75.5	87.9	88.0	96.1	76.4	80.1
	Pyramid [33]	79.0	89.0	88.2	95.7	76.9	78.9
	Relation-Net [16]	78.6	89.7	88.9	95.2	75.6	77.9
	RGA-SC [32]	-	-	88.4	96.1	77.4	81.1
	CDNet [9]	76.8	88.6	86.0	95.1	-	-
	CAL [18]	76.4	87.2	87.0	94.5	-	-
ViT	TransReID [5]	80.6	89.6	88.2	95.0	-	-
	DCAL [40]	80.1	89.0	87.5	94.7	-	-
	AAformer [41]	80.0	90.1	88.0	95.4	79.0	80.3
	PFD [27]	82.2	90.6	89.6	95.5	-	-
	CLIP-ReID [11]	82.5	90.0	89.6	95.5	80.3	81.6
LVLM	LVLM-ReID	82.8	92.2	89.2	95.6	82.3	84.6