Comparison of YOLOv8 and Detectron2 on Crowd Counting techniques

With the increase of security and surveillance cameras in public spaces has also created a demand for analyzing them and Crowd counting has gained a lot of popularity in recent times as it is viewed as a one of the major components for ensuring public safety and monitoring as it is a very laborious task for a person to sit in front of a security footage. It can be a very challenging task to create a dataset for training an algorithm and laborious data annotation is required for training and testing of the algorithms. This research paper discusses the 2 popular approaches of crowd counting in use today and examines their key features. The techniques used here are YOLOvS and Detectron2 which can be used for object and image detection and segmentation. YOLO stands for”You Only Look Once” and it is a popular algorithm as it is suitable for a wide variety of object detection, image segmentation and image classification tasks.Detectron2 is a Facebook AI Research (FAIR)’s library for object detection and segmentation in images. In our research, we found YOLOvS performance is better than detectron2. Detectron2 has been performed better than YOLOvS in terms of overall accuracy, but it has been observed on the basis of the experiments conducted above that YOLOvS is able to detect more objects than Detectron