Global Feature Learning with Human Body Region Guided for Person Re-identification
Person reidentification(re-id)is a very challenging task in video surveillance due to background clutters,variations in occlusion,and the human body misalignment in the detected images.To tackle these problems,we utilize a multi-channel convolutional neural network(CNN)with a novel embedding training strategy.First,some parts of the body were detected with existing methods of human pose estimation and then different parts were feed into different network branches to learn local and global representations.But for the global network branch,we proposed a embedding strategy for training,which uses local features to guide learning more robust global features.The promising experimental results on the large-scale Market-1501 and CUHK03 datasets demonstrate the effectiveness of our proposed embedding training strategy for features.
Person reidentification (re-id) Fusion strategy Sub-regions
Zhiqiang Li Nong Sang Kezhou Chen Chuchu Han Changxin Gao
Key Laboratory of Ministry of Education for Image Processing and Intelligent Control,School of Automation,Huazhong University of Science and Technology,Wuhan 430074,China
国际会议
广州
英文
15-25
2018-11-23（万方平台首次上网日期，不代表论文的发表时间）