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

为什么7x7的深度卷积可以分解成3x3和1x11和11x1的深度卷积? #3

Open
EchoXu98 opened this issue Apr 4, 2023 · 8 comments

Comments

@EchoXu98
Copy link

EchoXu98 commented Apr 4, 2023

您好!我想请问下,就是我标题中的问题“为什么7x7的深度卷积可以分解成3x3和1x11和11x1的深度卷积?” 这种分解是有严格数学推导吗?可以推导出分解前和分解后,二者的卷积核大小相同?还是因为效果好才这样分解?If you don't understand my question, I can ask it in English!

@yuweihao
Copy link
Collaborator

yuweihao commented Apr 4, 2023

Hi @EchoXu98 ,

Thanks for your attention. "decompose" here is more like "replace" from the aspect of performance. Depthwise convolution and inception depthwise convolution are not mathematically equivalent.

@sdu-huanhq
Copy link

sdu-huanhq commented Apr 20, 2023

您好,请问Inception dep. conv中的参数量(2k=9)C/8是怎么计算出来的?Inception dep conv的参数量应该和g的设置有关系吧。
Hello, may I ask how to calculate the parameter number (2k=9)C/8 in Inception dep. conv?

@yuweihao
Copy link
Collaborator

Hi @sdu-huanhq ,

Thanks for your attention. Yes, g is set as 1/8 by default in Table 1. Thanks for your reminder, I will add this information in the table caption in the next revision.

@sdu-huanhq
Copy link

Hi @sdu-huanhq ,

Thanks for your attention. Yes, g is set as 1/8 by default in Table 1. Thanks for your reminder, I will add this information in the table caption in the next revision.

Thanks for your answer and thumbs up for your work

@yuweihao
Copy link
Collaborator

Thanks for your answer and thumbs up for your work

Thank you so much!

@116022017144
Copy link

您好,请问Inception dep. conv中的参数量(2k=9)C/8是怎么计算出来的?Inception dep conv的参数量应该和g的设置有关系吧。 Hello, may I ask how to calculate the parameter number (2k=9)C/8 in Inception dep. conv?

I want to know where is the (2k+9) from?I know the C/8 is the parallel channel count

@yuweihao
Copy link
Collaborator

yuweihao commented May 8, 2023

您好,请问Inception dep. conv中的参数量(2k=9)C/8是怎么计算出来的?Inception dep conv的参数量应该和g的设置有关系吧。 Hello, may I ask how to calculate the parameter number (2k=9)C/8 in Inception dep. conv?

I want to know where is the (2k+9) from?I know the C/8 is the parallel channel count

2k from 1xk conv and kx1 conv and 9 from 3x3 conv.

@116022017144
Copy link

116022017144 commented May 8, 2023 via email

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

4 participants