Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.
And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.
The confusing positional arguments in the definition of focal loss
See original GitHub issueHi, I think you have a small but quite critical mistake in: https://github.com/qubvel/segmentation_models/blob/ba5cfca2b30be14050304096dba2fe0963b588e3/segmentation_models/losses.py#L238
In here, the arguments aboved are passed to the function below in the form of positional arguments: https://github.com/qubvel/segmentation_models/blob/ba5cfca2b30be14050304096dba2fe0963b588e3/segmentation_models/base/functional.py#L286
Here comes the confusing part. I think the value of alpha which we expect it to be, is replaced with the value of gamma. The same thing happens to gamma, too. But I’m not 100% certain about this.
If anyone sees this issue, I hope you can give a clear answer.
Issue Analytics
- State:
- Created 4 years ago
- Comments:9 (5 by maintainers)
Top Results From Across the Web
Top Related Medium Post
Top Related StackOverflow Question
Troubleshoot Live Code
Top Related Reddit Thread
Top Related Hackernoon Post
Top Related Tweet
Top Related Dev.to Post
Top Related Hashnode Post
Hi @LeeJayHui Thanks for pointing out about mistake in positional arguments! Can you create a PR to fix it? About
alphacoefficient… It not clear (for me) from paper is it should bealphaand(1 - alpha)or justalpha. Because in paper it is noted asalpha_twithout description.You are absolutely right, there are a lot of ways for improvement! If you have any ideas you can open a Pull Request and propose changes!