A question about generating eigencam for yolov5

[chakloong883]

I read the tutorial, and I want to ask a question about the generating a eigencam for yolov5. Why doesn't yolov5 need to write the target function when generating eigencam, but faster-rcnn needs it? Can I use the method of faster-rcnn to generate eigencam for yolov5? Which method is more accurate?
By the way, Can I use gradcam's method on yolov5? If used, whether a boundingbox can only correspond to a cam, can it only generate one cam like eigencam.