search

Qengineering / YoloV7-ncnn-Raspberry-Pi-4

YoloV7 for a bare Raspberry Pi using ncnn.
https://qengineering.eu/deep-learning-examples-on-raspberry-32-64-os.html
BSD 3-Clause "New" or "Revised" License
91 stars 19 forks source link

the model is yolov7-tiny, but the anchor size is yolov7? #2

Open hellocodelxy opened 1 year ago

hellocodelxy commented 1 year ago

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?

Qengineering commented 1 year ago

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

evanshlom commented 1 year ago

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

Qengineering commented 1 year ago

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.