Skip to content
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

测试帧率问题 #10

Closed
luckywangchenxi opened this issue Aug 25, 2021 · 7 comments
Closed

测试帧率问题 #10

luckywangchenxi opened this issue Aug 25, 2021 · 7 comments

Comments

@luckywangchenxi
Copy link

luckywangchenxi commented Aug 25, 2021

您好,复现shufflev2-yolov5代码与官方yolov5s代码对比,在同一实验环境下测试同一段视频为何测试速度反而是yolov5s更快一些(测试了两次)。

shufflev2-yolov5 yolov5s
96.353s 92.874s
95.978s 90.501s
@luckywangchenxi luckywangchenxi changed the title 复现帧率问题 测试帧率问题 Aug 25, 2021
@ppogg
Copy link
Owner

ppogg commented Aug 25, 2021

麻烦朋友告诉一下你测试时的硬件设备信息:比如架构是x86,使用GPU,显卡是多少这些~

@luckywangchenxi
Copy link
Author

luckywangchenxi commented Aug 25, 2021

您好,系统:Ubuntu 18.04.5 LTS (GNU/Linux 5.4.0-80-generic x86_64)
使用GPU
显卡:TIAN V
conda环境如下:
图片

@ppogg
Copy link
Owner

ppogg commented Aug 25, 2021

噢,我知道为什么了,shufflenet和mobilenet设计之初是为了适配芯片或者端侧设备,但是如果你使用GPU来跑这些网络或者包含这些boacbone的检测网络,效率会及其的慢,这主要是涉及到一个吞吐量和缓存的问题,在不同硬件设备上的推理时长我已进行过测试:
image

另外,如果朋友你想更加深入了解之所以会慢的原因,这里有一篇博客,讲的特别好:
为什么 MobileNet、ShuffleNet 在理论上速度很快,工程上并没有特别大的提升?

@ppogg
Copy link
Owner

ppogg commented Aug 25, 2021

如果你是学生,准备发论文的话,你可以尝试将你的视频在芯片上或这旧手机端进行检测,会有一个非常大的差距对比

@luckywangchenxi
Copy link
Author

好的,谢谢您的建议

@ppogg
Copy link
Owner

ppogg commented Aug 25, 2021

不客气,后期会尝试使用repvgg进行backbone的替换,同时兼容cpu和gpu,你可以持续关注这个repo,不定时更新

@luckywangchenxi
Copy link
Author

好的,一定

@ppogg ppogg closed this as completed Nov 30, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants