Abstract: In recent years, the use of object proposal as a preprocessing step for target detection to improve computational efficiency has become an effective method. Good object proposal methods should have high object detection recall rate and low computational cost, as well as good location accuracy and repeatability. However, it is difficult for current advanced algorithms to achieve a good balance in the above performance. Therefore, it is especially important to ensure that the recall rate and location quality are not degraded while accelerating object generation.For this problem, we propose a class-independent object proposal algorithm BIHL. It combines the advantages of window scoring and superpixel merging. First, a binarized horizontal high frequency component feature and a linear classifier are used to learn and generate a set of candidate boxs with a objective score. Then, the candidate boxs are merged based on the principle of location and score proximity. Different from superpixel merging algorithm, our method does not use pixel level operation to avoid a lot of computation without losing performance. Experimental results on the VOC2007 dataset and the VOC2007 synthetic interference dataset containing 297,120 test images show that when including difficult-to-identify objects with an IOU threshold of 0.5 and 10000 budget proposals, our method achieves a 99.3% detection recall and an mean average best overlap of 81.1% . The average processing time of our method for all test set images is 0.0015 seconds, which is nearly 3 times faster than the current fastest method. In repeatability testing, our method is the method with the highest average repeatability among the methods that achieve good repeatability to various disturbances, and the average repeatability is 10% higher than RPN. The code will be published in {https://github.com/JiangChao2009/BIHL}

Joint Boundary-Aware Method and Bayesian Inference for Object Proposal

Boundary-aware Box Refinement for Object Proposal Generation

Edge Boxes: Locating Object Proposals from Edges

Improved Saliency Detection Based on Bayesian Framework for Object Proposal.

Improved Proposal Generation Method Applying Bayesian Integration

Directional Edge Boxes: Exploiting Inner Normal Direction Cues for Effective Object Proposal Generation

BIHL:A Fast and High Performance Object Proposals based on Binarized HL Frequency

An Improved Approach for Object Proposals Generation

Object Proposal Generation with Fully Convolutional Networks

Object-Level Proposals.

Object Proposal on RGB-D Images Via Elastic Edge Boxes

A Minimum Barrier Distance Based Saliency Box for Object Proposals Generation

Adobe Boxes: Locating Object Proposals Using Object Adobes

Elastic Edge Boxes for Object Proposal on RGB-D Images.

Cascaded Sparse Spatial Bins for Efficient and Effective Generic Object Detection

Context-Aware Video Object Proposals.

Improving Object Proposals with Multi-Thresholding Straddling Expansion

Box Aggregation For Proposal Decimation: Last Mile Of Object Detection

Improving Object Proposals with Top-Down Cues

Deep Contour and Symmetry Scored Object Proposal.