Small Object Detection Using Context Information Fusion in Faster R-CNN

Citations Over Time

Abstract

Currently, most of the object detection research focuses on detecting a big object covering large part of the image. The problems of detecting the small object covering small part of the image are largely ignored. The difficulty of small object detection is that small objects have large quantity and less pixel (less information) and cover small part of the images. In this paper, we aim at improving the accuracy of small object detection. Firstly, we use a subset of the COCO [1] dataset to build a benchmark database. This benchmark database is specifically designed to evaluate the performance of object detection algorithms on small object detection. Secondly, we improve Faster R-CNN [2]. The improvements include a more flexible context information integration method. The experiments show that the improved Faster R-CNN algorithm has a good performance on the accuracy and recall rate of small object detection. Our small object detection algorithm is able to strike the balance between detection speed and detection accuracy.

Related Papers

• → Object Detection Algorithms for Video Surveillance Applications(2018)174 cited
• → A Survey on Object Detection Methods in Deep Learning(2021)17 cited
• → Object Detection using Deep Learning(2021)8 cited
• → MULTIPLE OBJECT DETECTION IN IMAGES USING YOLO(2022)
• → Object detection techniques for real-time applications(2022)