Weakly-supervised object detection via mining pseudo ground truth bounding-boxes

作者：

Highlights：

• A novel W2F framework for weakly-supervised object detection is proposed.

• The PGE algorithm is designed to mine the accurate pseudo ground truths.

• The PGA algorithm is proposed to refine pseudo ground truths.

• An IGL approach is proposed to further enhance quality of pseudo ground truths.

• The performance of weakly-supervised detection boosts a lot by using our method.

摘要

•A novel W2F framework for weakly-supervised object detection is proposed.•The PGE algorithm is designed to mine the accurate pseudo ground truths.•The PGA algorithm is proposed to refine pseudo ground truths.•An IGL approach is proposed to further enhance quality of pseudo ground truths.•The performance of weakly-supervised detection boosts a lot by using our method.

论文关键词：Weakly-supervised learning,Object detection,Pseudo ground truth,Iterative learning,Deep learning

论文评审过程：Received 23 January 2018, Revised 7 June 2018, Accepted 1 July 2018, Available online 5 July 2018, Version of Record 10 July 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.07.005