Bag of Tricks and A Strong Baseline for Deep Person Re-identification

17 Mar 2019  ·  Hao Luo, Youzhi Gu, Xingyu Liao, Shenqi Lai, Wei Jiang ·

This paper explores a simple and efficient baseline for person re-identification (ReID). Person re-identification (ReID) with deep neural networks has made progress and achieved high performance in recent years. However, many state-of-the-arts methods design complex network structure and concatenate multi-branch features. In the literature, some effective training tricks are briefly appeared in several papers or source codes. This paper will collect and evaluate these effective training tricks in person ReID. By combining these tricks together, the model achieves 94.5% rank-1 and 85.9% mAP on Market1501 with only using global features. Our codes and models are available at https://github.com/michuanhaohao/reid-strong-baseline.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Person Re-Identification DukeMTMC-reID BoT Baseline(RK) Rank-1 90.2 # 25
mAP 89.1 # 12
Person Re-Identification Market-1501 BoT Baseline(RK) Rank-1 95.43 # 36
mAP 94.24 # 12
Person Re-Identification Market-1501-C BoT (ResNet-50) Rank-1 27.05 # 22
mAP 8.42 # 21
mINP 0.20 # 21
Person Re-Identification MSMT17-C BoT (ResNet-50) Rank-1 20.20 # 4
mAP 5.28 # 4
mINP 0.07 # 3
Person Re-Identification UAV-Human Tricks mAP 63.41 # 1
Rank-1 62.48 # 2
Rank-5 84.38 # 2

Methods


No methods listed for this paper. Add relevant methods here