The Children and elderly are among the most vulnerable to accidents, in computer vision detection applications, they can be detected distinctly to avoid collision with them. In this case, the role of pre-trained algorithms come in to achieve this goal in real time, and in our work the most general detection algorithms were compared, and transferlearning techniques were studied and applied to train the detection model, we got distinct results in detection that reached 0.81 mAP in real time, using the Yolov5 model.

Deep Learning, Yolo, Pedestrian,Detection, Transfer Learning.