10000 Inverse class probability weights · Issue #85 · sacmehta/ESPNet · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Inverse class probability weights #85

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.

Sign up for GitHub

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

Jump to bottom
Open
ghost opened this issue Aug 18, 2022 · 0 comments
Open

Inverse class probability weights #85

ghost opened this issue Aug 18, 2022 · 0 comments

Comments

@ghost
Copy link
ghost commented Aug 18, 2022 •
edited by ghost
Loading

** also:
if class# range from 0 to 19 (20 classes), it would be hist = np.histogram(label_img, range=[0,self.classes-1],self.classes)[0]
**

I think there are issues with the calculation of the class weights?

In train/loadData.py, I don't think "hist = np.histogram(label_img, self.classes)" has the desired result.
In the absence of range, np.histogram will use the min and max value of the array...

Should it not be "hist = np.delete(np.histogram(label_img, range=[0,self.classes],self.classes+1)[0],[0])" ?
(i.e. for 20 classes, range is [0,20] with 21 bins, and bin "0" gets discarded)

I'd be glad to understand why I'm wrong otherwise :)

Nicolas.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
0 participants
0