A New Architecture of Feature Pyramid Network for Object Detection
Yichen Zhang, Jeong Hoon Han, Yong Woo Kwon, Young Shik Moon
Abstract
In recent years, object detectors generally use the feature pyramid network (FPN) to solve the problem of scale variation in object detection. In this paper, we propose a new architecture of feature pyramid network which combines a top-down feature pyramid network and a bottom-up feature pyramid network. The main contributions of the proposed method are two-fold: (1) We design a more complex feature pyramid network to get the feature maps for object detection. (2) By combining these two architectures, we can get the feature maps with richer semantic information to solve the problem of scale variation better. The proposed method experiments on PASCAL VOC2007 dataset. Experimental results show that the proposed method can improve the accuracy of detectors using the FPN by about 1.67%.