Vehicle Recognition Based on Improved Faster R-CNN

Abstract

The objects detected by the existing object detection algorithms are bounding boxes with a certain background, and the bounding boxes cannot be aligned with the objects. In order to solve the pixel alignment problem and improve the average accuracy, this paper introduces a mask based on the Faster R-CNN algorithm, which can extract the fine spatial layout of the target, so as to distinguish the target from the background and improve the average accuracy. For vehicle recognition, the improved model is trained on the vehicle attribute classification dataset, and the average accuracy of classification recognition is significantly improved. The vehicle recognition algorithm proposed in this paper is based on the improvement of Faster R-CNN, and the effectiveness of the method is verified by testing with the COCO dataset.