https://github.com/ultralytics/yolov5/releases/tag/v7.0
I downloaded the pre-trained model from the link above and obtained yolov5n.pt, then exported it to a TFLite model using the following command:
python export.py --weights yolov5n.pt --imgsz 416 --include tflite --int8
After that, I wrote code for real-time inference on the i.MX 8M PLUS EVK board by referring to Ultralytics' detection.py.
The problem is that YOLOv4-tiny (416x416) produced at least 10 fps regardless of resolution, but YOLOv5 (416 x 416) gives only about 2-3 fps starting from 320x240.
So I looked for example code in the NXP community, and it seems possible to request a test code.
Could I receive a test code?
