Improving object proposals with top-down cues

作者：

Highlights：

• We improve the ranking of object proposals by combining bottom-up cues and top-down generic object information.

• We propose a simple and effective strategy to mining useful top-down cues.

• We achieve higher recall with IoU = 0.8 compared with the state-of-the-art on the Pascal VOC2007 dataset.

摘要

Highlights•We improve the ranking of object proposals by combining bottom-up cues and top-down generic object information.•We propose a simple and effective strategy to mining useful top-down cues.•We achieve higher recall with IoU = 0.8 compared with the state-of-the-art on the Pascal VOC2007 dataset.

论文关键词：Object proposals,Object detection,Object recognition

论文评审过程：Received 11 August 2016, Revised 7 February 2017, Accepted 10 April 2017, Available online 13 April 2017, Version of Record 25 April 2017.

论文官网地址：https://doi.org/10.1016/j.image.2017.04.006