Selective Weakly Supervised Human Detection under Arbitrary Poses

作者：

Highlights：

• We propose a novel Selective Weakly Supervised Detection method which outperforms the previous state-of-the-art methods.

• We annotate a new large-scale data set called LSP/MPII-MPHB (Multiple Poses Human Body) for human body detection.

• We identify an easily ignored pitfall of the Noisy-OR model in MIL, which can significantly reduce the training efficiency of a MIL algorithm.

• We present a comprehensive and in-depth empirical study of the weakly supervised MIL method.

摘要

Highlights•We propose a novel Selective Weakly Supervised Detection method which outperforms the previous state-of-the-art methods.•We annotate a new large-scale data set called LSP/MPII-MPHB (Multiple Poses Human Body) for human body detection.•We identify an easily ignored pitfall of the Noisy-OR model in MIL, which can significantly reduce the training efficiency of a MIL algorithm.•We present a comprehensive and in-depth empirical study of the weakly supervised MIL method.

论文关键词：Weakly supervised learning,Human detection,Selective Weakly Supervised Detection (SWSD),Multi-instance learning (MIL)

论文评审过程：Received 19 June 2016, Revised 18 October 2016, Accepted 24 December 2016, Available online 28 December 2016, Version of Record 5 January 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.12.025