Agile Modeling Workflow
1
Compute Embeddings
Coming soon...
Provide a path to your ARU recordings in a public Google Storage bucket.gs://chirp-public-bucket/soundscapes/high_sierras/audio/*2
Search Recordings
Search through ARU recordings for species of interest using the computed embeddings. Provide a list of 6-letter eBird codes and let the app use Xenocanto recordings to search for similar recordings from your data set.
3
Annotate Recordings
Review and annotate 10-20 examples per class, distinguishing between song/call types.
4
Train & Classify
Train a classifier on labeled outputs and use it to classify all recordings.
5
Examine Classifier Outputs
Review the outputs from the classifier and label any misclassifications.