A small object detection algorithm based on improved faster RCNN

Abstract

Faster RCNN, as a classical detection algorithm, still faces a huge challenge in detecting small objects. Therefore, we introduce a multi-scale auxiliary feature fusion strategy to make sure that each layer of features contains rich semantic and spatial information. Firstly, we introduce shallow features extracted by a multi-scale auxiliary feature network into the backbone network, as a way to ensure that there is sufficient spatial information for detecting small objects even for the deepest feature. Secondly, we design a fusion module to fuse the auxiliary feature and backbone feature. Finally, to make the object proposal boxes positioning more precise in the ROI classification and regression network, replace RoIPool with RoIAlign. Our experiments are conducted on PASCAL VOC and KITTI autopilot datasets. Compared with the conventional methods, the improved Faster RCNN algorithm has 2.48% and 3.09% improved in mean average precision on PASCAL VOC and KITTI datasets, respectively.