Evaluating the Performance of the YOLO Object Detection Framework on COCO Dataset and Real-World Scenarios

Object detection is one of the cutting-edge tools of computer vision to present the content of an image or video frame. Object recognition describes the entire scene in the frame. So many sophisticated algorithms are available to detect and recognize the object in the frame. The YOLO framework has changed the world of computer vision because of its accurate detection feature. The proposed article focuses on the YOLO framework for object detection. The YOLO framework has been applied to the various images of the COCO dataset. The presence of object and object class probability is measured in the form of an intersection of unions. The YOLO framework is applied to real-time input sources like groups of people as multiple objects for validation. It gives an accuracy between 73% and 85% for the group of people as multiple objects.