Boundary-aware box refinement for object proposal generation

Object proposals have been widely used in object detection to speed up object searching. However, many of existing object proposal generators have pool localization quality, which weakens the performance of object detectors. In this paper, we present an effective approach to improve the localization quality of object proposals. We leverage the boundary-preserving property of superpixels and design an efficient algorithm for object proposal refinement. Our approach first performs bounding box alignment to adapt proposals to potential object boundaries, and then diversifies the proposals via multi-thresholding superpixel merging. The algorithm only takes 0.15 s and can be applied to any existing proposal methods to improve their localization quality. Extensive experiments on PASCAL VOC 2007 and ILSVRC 2013 datasets show our approach significantly and consistently improves the recall, localization accuracy, and detection performance of existing proposal methods. When combining with Region Proposal Network, our approach outperforms the state-of-the-art object detectors by a large margin.