Object detector with enriched global context information

  • PDF / 2,402,270 Bytes
  • 21 Pages / 439.642 x 666.49 pts Page_size
  • 97 Downloads / 258 Views

DOWNLOAD

REPORT


Object detector with enriched global context information Jingjuan Guo1,2 · Caihong Yuan3 · Zhiqiang Zhao2 · Ping Feng4 · Yihao Luo1 · Tianjiang Wang1 Received: 1 October 2019 / Revised: 10 July 2020 / Accepted: 29 July 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020

Abstract How to add more context information and bring more accurate detection is an important problem to be considered in object detection. In this paper, we propose a new object detector with enriched global context information by a pyramid feature pool module and several global activation blocks, named EGCI-Net, which is a one-stage object detector from scratch as DSOD.The global activation blocks are added into the backbone sub network of the detector to weaken the local information of the detected object feature maps and increase the global context of them. And the pyramid feature pool module produces multi-scale global context features to supervise the pyramid features by multi-scale global average pooling. Then the features obtained by the main structure are fused with the pyramid pooling features to merge into the final multibox detector. We have evaluated our detector on the Pascal VOC  Jingjuan Guo

jj [email protected]  Tianjiang Wang

tjwang [email protected] Caihong Yuan [email protected] Zhiqiang Zhao zq [email protected] Ping Feng fengping [email protected] Yihao Luo [email protected] 1

School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074, China

2

School of Information Science and Technology, Jiujiang University, Jiujiang 332005, China

3

School of Computer and Information Engineering, Henan University, Kaifeng 475004, China

4

International Joint Research Center For Data Science and High-Performance Computing, Guizhou University of Finance and Economics, Guiyang, 550025, China

Multimedia Tools and Applications

and MS COCO datasets. The experimental results show that our proposed detector achieves better results than DSOD and exceeds most of the existing excellent detectors, especially detects partially occluded objects and small objects well. Keywords Object detection · Global context information · Pyramid pooling features

1 Introduction Object detection is an important direction in image processing. It can be applied to face detection [45], vehicle detection [2], pedestrian detection [22] [42] and other practical applications [40] [33] [50] [21]. At present, object detection based on deep learning is divided into two major directions, a two-stage detector such as R-CNN [9], Fast R-CNN [8], Faster R-CNN [32], R-FCN [5] and a one-stage detector such as SSD [27], YOLO [30]. The detection speed of the two-stage detector is slightly slower but the detection accuracy is higher. And the one-stage detector is just the opposite, faster but slightly less effective. Among them, the one-stage detector is attracting more attention because the detection speed is faster and it is more likely to be applied to practical applications. However, most of detector network a