Open angelogeninatti opened 5 months ago
This is very interesting use case. What we should do is to add more modalities (e.g., Audio) into our framework similarly to others, instead of concatenate different modalities into one. By default, any modality should extend from and behave like a FeatureModality.
Description
In my use case, I need to incorporate multiple modalities inside my data, including Audio and Video. As of right now, the only way I found to incorporate them both is to concatenate them in a single FeatureModality object, which is not ideal. It would be best to be able to use a list of FeatureModality objects.
Expected behavior with the suggested feature
As of right now, the only way to do it is: item_embeddings = FeatureModality(features=[Concatenated Modalities], ids=...) ratio_split = RatioSplit( data=dataset, test_size=0.1, item_feature=item_embeddings )
It would be best to have something like: item_embeddings = [FeatureModality(features=Modality 1, ids=...), FeatureModality(features=Modality 2, ids=...)] ratio_split = RatioSplit( data=dataset, test_size=0.1, item_feature_list=item_embeddings )
Other Comments
None