Improvement of Object Detection Based on Faster R-CNN and YOLO

The development of artificial intelligence technology has been greatly assisted by object detection. The object detector like you-only-look-once (YOLO) v2 can detect an object in real-time and also with good accuracy. However, except for the lower computation cost and faster speed, the single-stage detector YOLO v2 is not as good as the two-stage detectors like Faster R-CNN in terms of accuracy; more improvement is needed to increase the accuracy. This paper uses the Kalman filter to fuse Faster R-CNN and YOLO v2 to obtain better detection accuracy. The results from Faster R-CNN are served as observation due to its better accuracy, while that from YOLO v2 as state variables. Experiment is carried out in video samples containing vehicle images. The results show that the fusion of the two algorithms by using the Kalman filter can provide better object detection.