Extraction plugin for Hansken to classify and segment audio, relying heavily on the model in inaSpeechSegmenter
Please note that this model was not developed or trained by the Netherlands Forensic Institute. This is merely a container that supports running the classification on Hansken projects.
This Docker container on its own is not that useful. It expects to be called by Hansken's extraction service, and writes its results to a Hansken project.
Build the Docker image:
docker build -t audiosegmentation .
When developing and debugging, it's possible to avoid the Docker build step by simulating a tool run using hansken.py
Install the dependencies from setup.py
and run the main entry point in audio_segmentation.py
, specifying the required
command line arguments.