A novel efficient method for extraction of object proposals is introduced. Its "objectness" function exploits deep spatial pyramid features, a new fast-to-compute HoG-based edge statistic and the EdgeBoxes score. The efficiency is achieved by the use of spatial bins in combination with sparsity-inducing group normalized SVM. State-of-the-art recall performance is achieved on Pascal VOC07, significantly outperforming methods with comparable speed. Interestingly, when only 100 proposals per image are considered our method attains 78% recall on VOC07. The novel method improves mAP of the RCNN state-of-the-art class-specific detector, increasing it by 10 points when only 50 proposals are utilized in each image. Despite training the system on twenty classes, results on two-hundred class ILSVRC2013 dataset confirm generalization to previously unseen data.

本研究提出了一种高效的目标提取方法，通过利用深度空间金字塔特征、基于HoG的新型快速计算边缘统计和EdgeBoxes得分的'objectness'函数，以及与稀疏感知组归一化SVM结合的新型空间分bin方式，达到了与同等速度方法相比更高的Pascal VOC07的回收性能，甚至仅考虑每个图像100个提议时，方法仍达到了78％的召回率，并且在使用每个图像中仅50个提议时，改善了RCNN最先进的类别特定检测器的mAP，提高了10个百分点。该系统对于全球200个类别在ILSVRC2013数据集中的检测表现良好，证实了其泛化能力。

级联稀疏空间块用于高效有效的通用物体检测