These are the model weights trained as reported in the paper.
They are configured with the SE-Resnet34 backbone, followed by linear heads corresponding to the approaches used.
Unzip and paste the folders into src/results/ (refer to the hierarchy example below).
You can use the evaluation command (at README.md/#3-3) to see how it works.
/
├─ src\
├─ results\
├─ DCASE2020-ADYOLO\
├─ DCASE2021-ADYOLO\
...
└─ DCASE2022-ADYOLO\