Qengineering / YoloV7-ncnn-Raspberry-Pi-4

YoloV7 for a bare Raspberry Pi using ncnn.

Home Page:https://qengineering.eu/deep-learning-examples-on-raspberry-32-64-os.html

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

the model is yolov7-tiny, but the anchor size is yolov7?

hellocodelxy opened this issue · comments

hi, i have a question.
your model is yolov7-tiny, but he anchor size in yolov7.cpp is yolov7's size. Is there any error?

Very sharp.
It doesn't matter that much. If you compare the outcome, the YoloV7 anchors are even better positioned than the tiny counterparts.
image
image

Does this mean yolov7 is being used instead of yolov7-tiny?

Dear @evanshlom,

No. YoloV7-tiny is still used. Hence the impressive inference times.
Only the algorithm finding the boxes is feed by YoloV7 parameters.
Which, surprisingly, gives better results in this situation than the tiny anchors.