Person Re-Identification Based on Multi-Layer Feature

Kewen Liu; Panpan Fang; Hongxia Xiong; Chaoyang Liu; Yuan Ma; Xiaojun Li; Yalei Chen

doi:10.3788/LOP57.081503

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 8 >Page 081503 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 8, 081503 (2020)

Person Re-Identification Based on Multi-Layer Feature

Kewen Liu^1、2, Panpan Fang^1、2, Hongxia Xiong^3、*, Chaoyang Liu⁴, Yuan Ma^1、2, Xiaojun Li^1、2, and Yalei Chen^1、2

Author Affiliations

¹School of Information Engineering, Wuhan University of Technology, Wuhan, Hubei 430070, China

²Hubei Key Laboratory of Broadband Wireless Communication and Sensor Networks, Wuhan University of Technology, Wuhan, Hubei 430070, China

³School of Civil Engineering and Architecture, Wuhan University of Technology, Wuhan, Hubei 430070, China

⁴State Key Laboratory of Magnetic Resonance and Atomic Molecular Physics, Wuhan Institute of Physics and Mathematics, Chinese Academy of Sciences, Wuhan, Hubei 430071, China

show less

DOI: 10.3788/LOP57.081503 Cite this Article Set citation alerts

Kewen Liu, Panpan Fang, Hongxia Xiong, Chaoyang Liu, Yuan Ma, Xiaojun Li, Yalei Chen. Person Re-Identification Based on Multi-Layer Feature[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081503 Copy Citation Text

show less

Fig. 1. Residual module

Download full size

Fig. 2. Network structure

Download full size

Fig. 3. Graph of mapping function

Download full size

Fig. 4. Rank-1 results of Market-1501 and DukeMTMC-ReID. (a) Market-1501; (b) DukeMTMC-ReID

Download full size

Configuration	Parameter
Operating system	Windows 7,64 bit
RAM(random access memory)	8G
CPU (central processing unit)	Intel PentiumG3260 3.3 GHz
GPU (graphics processing unit)	NVIDIAGTX 1070Ti
Software platform	Pytorch 1.1

Table 1. Simulation platform parameters

Stage	Rank-1 /%	mAP /%
Stage 1	61.0	69.9
Stage 2	81.8	71.3
Stage 3	89.1	74.9
Stage 4	88.7	73.2

Table 2. Simulation results of different stages

Method No.	Rank-1 /%	mAP /%
1	61.0	69.9
2	83.4	72.6
3	89.7	77.9
4	90.4	84.6
5	89.8	83.7

Table 3. Simulation results of different methods

Method	Market-1501		DukeMTMC-ReID
Method	Rank-1	mAP	Rank-1	mAP
Method 4	90.4	84.6	84.6	80.2
Method 5	89.8	83.7	84.4	80.0
Propoed method	91.7	86.8	84.9	80.7

Table 4. Simulation results of multi-layer feature fusion%

Method	Time /s
Method	Test time	Per match(19732)
GLAD^[12]	368	0.0186
ThriNet^[19]	343	0.0173
DaRe^[21]	315	0.0159
Proposed method	328	0.0166

Table 5. Comparative results of running time of different methods

Method	Market-1501		DukeMTMC-ReID
Method	Rank-1	mAP	Rank-1	mAP
SVDNet^[22]	82.3	62.1	76.7	56.8
MultiScale^[23]	88.9	73.1	79.2	60.6
GLAD^[12]	89.9	73.9	-	-
ThriNet^[19]	84.9	69.1	-	-
DaRe^[21]	90.8	85.9	84.4	79.6
Proposedmethod	91.7	86.8	84.9	80.7

Table 6. Performance comparison between proposed algorithm and mainstream methods%

Kewen Liu, Panpan Fang, Hongxia Xiong, Chaoyang Liu, Yuan Ma, Xiaojun Li, Yalei Chen. Person Re-Identification Based on Multi-Layer Feature[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081503

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information