SFGNet detecting objects via spatial fine-grained feature and enhanced RPN with spatial context

Object detection, which is one of the most fundamental visual recognition tasks, has been a hotspot in computer vision. CNN (Convolutional Neural Networks) have widely employed for building detector. Due to success RPN (Region Proposal Network), two-stage detectors get both classification accuracy and precise regression bounding boxes. However, they still struggle small-size object detection. In this paper, we present deep network, namely Spatial Fine-Grained Network (SFGN). The SFGN that exploits Features (SFGF) concatenates higher resolution features, fine-grained with low features high-level semantic by stacking spatial features. An enhanced region proposal generator proposed objectless small obtain set proposal. contextual information surrounding interest embedded using local increasing useful discriminating background. For improving detection performance, use simple yet surprisingly effective online hard example mining (OHEM) algorithm training generator. It embeds an efficiently implemented soft non-maximum suppression (soft-NMS) replacing tradition NMS consistent improvements without computational complexity inference. On PASCAL VOC 2007 2012 datasets, our improves baseline model from 81.2% mAP 80.6% mAP. MS COCO dataset, also performs better than model. As intuition suggests, results provide strong evidence accuracy, especially test.