New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Replace Conv2d in my network, however it becomes slower, why? #26
Comments
More details? Which toolkit/hardware did you use? We test GhostNet on ARM with TFLite. |
Thanks for your reply! |
Same question,I tested under the Darknet framework, and it will be faster when using the CPU, but it will be much slower when using the GPU. |
Your observation is normal. Ghost module is more suitable for ARM/CPU, and not friendly for GPU due to the Depthwise Conv. |
thanks! |
Thank you for your answer. After using GhostMoudle, I found that the detection speed has slowed down. This has troubled me for a long time. |
Above all, thanks for your great work! It really inspires me a lot! But now I have a question.
I replace all the
Conv2d
operations in my network except the final ones, the model parameters really becomes much more less.However, when testing, I found that the average forward time decreases a lot by the replacement (from 428FPS down to 354FPS).
So, is this a normal phenomenon? Or is this because of the
concat
operation?The text was updated successfully, but these errors were encountered: