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

About multi target per image #1

Open
wanghr-git opened this issue Jun 5, 2022 · 4 comments
Open

About multi target per image #1

wanghr-git opened this issue Jun 5, 2022 · 4 comments

Comments

@wanghr-git
Copy link

Hi, I have a lot of admiration for your excellent work. Notice that in the images in the example in your article, each image has only one target, have you tried the situation where there are multiple targets in one image, how is the effect? Will multiple targets affect the "Boundary Key-Point Map Generator"?

@wanghr-git wanghr-git reopened this Jun 5, 2022
@jcwang123
Copy link
Owner

Hi, thanks for your appreciation.
Actually, this code is designed for targets with a single class only.
However, it can be re-produced for multi-class segmentation by:
(1) generating the map for each class and merging all maps into one for attention and boundary-aware supervision.
(2) changing the prediction head

@wanghr-git
Copy link
Author

Hi, thanks for your appreciation. Actually, this code is designed for targets with a single class only. However, it can be re-produced for multi-class segmentation by: (1) generating the map for each class and merging all maps into one for attention and boundary-aware supervision. (2) changing the prediction head

Thank you for your reply.Sorry for not expressing my meaning well. In fact, what I mean is that there is only one object region per image in your dataset. If there are multiple objects of the same class in an image, will your model be affected, e.g.
image

@jcwang123
Copy link
Owner

hah, I got it.
In this case, the generation will be affected. But the major goal to find the most ambiguous points through the segmentation map can be borrowed, i.e., the corners in this map you present. It may be helpful~

@wanghr-git
Copy link
Author

hah, I got it. In this case, the generation will be affected. But the major goal to find the most ambiguous points through the segmentation map can be borrowed, i.e., the corners in this map you present. It may be helpful~

Thank you so much for such a quick reply, my problem has been solved.Best wishes!

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