Tool for exploring and improving AI detection of SRKW calls to generate expert-labeled data -- prototype at https://ai4orcas.net/portfolio/pod-cast-annotation-system/
MIT License
5
stars
1
forks
source link
Script for creating new Pod.Cast annotation "rounds" #7
Simplify the process of creating new annotation "rounds".
@scottveirs maintains Orcasound hydrophone events in the Gsheet over time here -- and there's a column for prioritizing which events are optimal for annotation rounds in Pod.Cast | Help listen for Orcas!.
Creating a new round involves the following:
Pulling audio files from Orcasound AWS S3 and breaking them into 1 min *.wav chunks
Processing these *.wav chunks with a model and save predictions in corresponding JSON files
(optional) filter out very low confidence chunks to reduce annotation burden
Uploading these files in appropriate structure to Azure Blob
No update is required to the flask web-app, things will be available under the /roundN suffix to the URL
Identifying interesting events from the GSheet is OK to trigger manually now - this issue is for a prepare_for_podcast.py script to simplify the rest of it.
Simplify the process of creating new annotation "rounds".
@scottveirs maintains Orcasound hydrophone events in the Gsheet over time here -- and there's a column for prioritizing which events are optimal for annotation rounds in Pod.Cast | Help listen for Orcas!.
Creating a new round involves the following:
/roundN
suffix to the URLIdentifying interesting events from the GSheet is OK to trigger manually now - this issue is for a
prepare_for_podcast.py
script to simplify the rest of it.