Insights into Soundscape Synthesis and Energy consumption of Sound Event Detection systems

Francesca Ronchini (Politecnico di Milano) joins us to discuss her PhD research:

The main objective of a sound event detection (SED) system is to detect and classify events present in an audio clip. DCASE Task 4 aims to develop a SED system using a heterogeneous dataset containing both real recorded sounds and synthetically generated soundscapes. In this study, particular attention is given to studying the effect of non-target events present in the synthetic soundscapes on a collection of state-of-the-art submissions from the DCASE Task 4 challenge. Additionally, an energy consumption metric will be introduced to create awareness about the ecological impact of training sound event detectors.
NOTE: this event is part of the DL4MIR workshop series (ccrma-mir.github.io); guest speaker talks are open to the broader CCRMA community.