The speed of the tiny model. about yolov7 HOT 5 OPEN

Thanks for the answer, When tested on my CPU, the difference between a YOLOv5-N and YOLOv5-S was approximately 100ms, which is really important. By 0.1 you mean 0.1second?

0.1ms

from yolov7.

Inference time CPU is usually proportional to FLOPs. And it seems very tiny model will reach the speed bottleneck on GPU, you could find that difference of inference time between YOLOv5-N and YOLOv5-S reported by YOLOv5 (r6.1) is only 0.1.

from yolov7.

Thanks for the answer,
When tested on my CPU, the difference between a YOLOv5-N and YOLOv5-S was approximately 100ms, which is really important.
By 0.1 you mean 0.1second?

from yolov7.

@JiaPai12138 Thanks!

from yolov7.

@JiaPai12138 Thanks!
np

from yolov7.