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

The meaning of lighting augmentation #51

Closed
voldemortX opened this issue May 29, 2021 · 4 comments
Closed

The meaning of lighting augmentation #51

voldemortX opened this issue May 29, 2021 · 4 comments

Comments

@voldemortX
Copy link

voldemortX commented May 29, 2021

Hi! I'm trying to understand what is the lighting augmentation? Are the eigen values representative of a light source & do I need to compute them for other datasets?

Could you help me with this @liuruijin17, thanks very much!

@liuruijin17
Copy link
Owner

The lighting functions come from CornerNet. And in their paper, they said "we adopt standard data augmentation including ..., random color jittering. Finally, we apply PCA to the input image." So the lighting augmentation here is to highlight the network the most principal image features.

You could compute them from other datasets, which would be more reasonable since not all datasets share the same intensity distribution.

@voldemortX
Copy link
Author

Thanks! So the numbers from your repo at the moment are the ones from COCO dataset?

@liuruijin17
Copy link
Owner

You're welcome.

Maybe. There are no comments from CornerNet's authors.

This question also shows up in CornerNet's repo, such as
issue 163, 85, 75, but no responses.

@voldemortX
Copy link
Author

@liuruijin17 Yes, I see that. Since princeton-vl/CornerNet#163 is open, I'll close this one and post future inquiries there.

voldemortX added a commit to voldemortX/pytorch-auto-drive that referenced this issue Jun 2, 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