A repo designed to convert audio-based "weak" labels to "strong" intraclip labels. Provides a pipeline to compare automated moment-to-moment labels to human labels. Methods range from DSP based foreground-background separation, cross-correlation based template matching, as well as bird presence sound event detection deep learning models!
I decided to use more e4e dedup syntax after discovering matrix missing, not making a build command, since python-semantic-versioning is containerized, the build command fails to run in python 3.9/3.10
Please resolve lock file conflicts