Open atahansrc1 opened 1 year ago
It depend on what device your used and what arguments you applied. For example, if you used CPU, inference speed mainly corresponds to FLOPs. If you used GPU without tensor core, due to yolov7 default using half precision inference, it will cause some overhead.
By the way, from yolov5 reported performance, we could find that yolov5s and yolov5n have almost same inference speed. It obviously does not make sense. However, due to we could not reproduce such faster inference speed of yolov5l and yolov5x as yolov5 official reported. We can only report all of results as what yolov5 official report.
Hi, I'm a beginner in these algorithms, and I'm sorry if this is a stupid question. I have a license plate dataset that includes 1200 images, and I trained both YOLOv5s and YOLOv7 with that dataset. When I compared YOLOv5s and YOLOv7, it appeared that YOLOv5s was faster than the YOLOv7. YOLOv5s processed the image at 100ms average. However, YOLOv7 processed the image at 350ms average. Is it normal? Thank you.