How can I compare the result of SSD and YOLO with mAP metric in hand object detection? How can I get mAP value of these two models?