-
Notifications
You must be signed in to change notification settings - Fork 4
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
Performance degradation compared to the previous psds eval #4
Comments
Hi, I have tested sed_scores_eval on the teacher scores you provided as follows
which gives a psds of 0.5054 as in you screenshot. Running the reference implementation based on psds_eval (that is used for testing) as follows
gives a psds_ref of 0.4956 which doesn't match your reported value. I am wondering if the reason for the difference might be in the way you are performing the decoding before computing psds from operating points. In your code (that I cannot execute btw) I see that you use Can you maybe provide a function (incl dependencies so I can execute it) where your reported results are both generated from the same scores that you provided? |
I use same code at dcase baseline and FDY-SED github (https://github.com/frednam93/FDY-SED) this used codes
|
Hi, so if you use different methods for post-processing you cannot expect the results to be comparable. If you think, however, the two functions are doing the same post-processing, then please provide a minimal working example where scores with the same post-processing give a psds1_2023 that is lower than psds1_2022. Otherwise I cannot see if there really is an issue here. |
Hello.
Problem :
Although this library was used, there is a problem that performance is lower in certain models compared to the existing psds score.
Please refer to the picture below for performance.
For performance implementation, related codes and materials are shared through the link below.
code for evaluation
scores : https://drive.google.com/file/d/1i_OPaFSQKnZZH6kQZr5cqxTdNsapn0mh/view?usp=share_link
The text was updated successfully, but these errors were encountered: