Sequential End-to-end Network for Efficient Person Search

18 Mar 2021  ·  Zhengjia Li, Duoqian Miao ·

Person search aims at jointly solving Person Detection and Person Re-identification (re-ID). Existing works have designed end-to-end networks based on Faster R-CNN. However, due to the parallel structure of Faster R-CNN, the extracted features come from the low-quality proposals generated by the Region Proposal Network, rather than the detected high-quality bounding boxes. Person search is a fine-grained task and such inferior features will significantly reduce re-ID performance. To address this issue, we propose a Sequential End-to-end Network (SeqNet) to extract superior features. In SeqNet, detection and re-ID are considered as a progressive process and tackled with two sub-networks sequentially. In addition, we design a robust Context Bipartite Graph Matching (CBGM) algorithm to effectively employ context information as an important complementary cue for person matching. Extensive experiments on two widely used person search benchmarks, CUHK-SYSU and PRW, have shown that our method achieves state-of-the-art results. Also, our model runs at 11.5 fps on a single GPU and can be integrated into the existing end-to-end framework easily.

PDF Abstract

Datasets


Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Person Search CUHK-SYSU NAE+SeqNet+CBGM MAP 94.8 # 7
Top-1 95.7 # 7
Person Search CUHK-SYSU OIM+SeqNet MAP 93.4 # 12
Top-1 94.1 # 13
Person Search CUHK-SYSU NAE+SeqNet MAP 93.8 # 11
Top-1 94.6 # 10
Person Search CUHK-SYSU OIM+SeqNet+CBGM MAP 94.3 # 9
Top-1 95.0 # 9
Person Search PRW OIM+SeqNet+CBGM mAP 46.6 # 11
Top-1 84.9 # 6
Person Search PRW OIM+SeqNet mAP 45.8 # 14
Top-1 81.7 # 13
Person Search PRW NAE+SeqNet mAP 46.7 # 9
Top-1 83.4 # 9
Person Search PRW NAE+SeqNet+CBGM mAP 47.6 # 7
Top-1 87.6 # 3

Methods