DCASE 2023 Workshop
Spotlight presentations of our papers in the DCASE 2023 Workshop (links to YouTube DCASE channel)
Manu Harju: Evaluating Classification Systems Against Soft Labels with Fuzzy Precision and Recall
James Afolaranmi: Sound Event Classification with Object-Based Labels
Irene Martin Morato: Aggregate or Separate: Learning from Multi-Annotator Noisy Labels for Best Classification Performance
Presenting results the proper way
When presenting the experimental results in a paper, confidence intervals provide a good way to demonstrate that the obtained improvement is really an improvement. Calculating confidence intervals is not that complicated, so why are they not commonly seen in our area of research? This is a small attempt to encourage our group to use.
Slides: Calculating confidence intervals