Multi-label learning of part detectors for occluded pedestrian detection
作者：
Highlights：
• We propose a multi-label learning approach to jointly learn part detectors which share decision trees to exploit correlations among parts and also reduce the computational cost of applying these part detectors.
• We explore several integration methods to integrate the part detectors learned by the proposed approach for occlusion handling.
• We exploit context to further improve the performance for pedestrian detection.
摘要
•We propose a multi-label learning approach to jointly learn part detectors which share decision trees to exploit correlations among parts and also reduce the computational cost of applying these part detectors.•We explore several integration methods to integrate the part detectors learned by the proposed approach for occlusion handling.•We exploit context to further improve the performance for pedestrian detection.
论文关键词：Pedestrian detection,Part detectors,Multi-label learning,Occlusion handling,Detector integration,Context
论文评审过程：Received 29 March 2018, Revised 30 June 2018, Accepted 27 August 2018, Available online 5 September 2018, Version of Record 17 September 2018.
论文官网地址：https://doi.org/10.1016/j.patcog.2018.08.018
