Closed Maxime-Bru closed 1 month ago
Hi @Maxime-Bru , this feature is implemented in the develop
branch as CNN.embed()
and will be included in the upcoming 0.11.0 release. To accomplish something similar in current/previous releases follow this example from the cookbook.
Thank you, this works perfectly!
I trained a custom CNN model following the tutorial (https://opensoundscape.org/en/latest/tutorials/train_cnn.html).
Now my model has a predict() method to get the prediction scores. I don't think it is yet possible to retrieve the embeddings: would it be possible to add a generate_embeddings() method to achieve this?