Crowd Abnormal Behaviour Detection and Comparative Analysis using YOLO Network

In the growing age, with the increase in traffic the number of accidents has also increased exponentially.A number of such cases can be seen in our daily life. There are several single stage detection algorithms available those are seriously concerned with the speed. Here, YOLO plays a major role in object detection due of its strong feature extracting and adaptive ability. This paper proposes a comparison between YOLOv3 and YOLOv3 tiny in which we observed that the accuracy of YOLOv3 algorithm is high with a requirement of higher detection time whereas the precision in both the cases is approximately equal. The maximum accuracy attained by YOLOv3 is 0.99 whereas in case of YOLOv3 tiny it is 0.75.