Justin Salamon

Nice progress @iranroman! @iranroman @magdalenafuentes excuse typos, writing on mobile. IIUC the events are annotated as time series of active/silent with a 100ms hop size. Is that correct? We are...

Hey, just catching up now. My first question is whether it makes sense to create a new annotation type over expanding the existing `Event` type. For example, the spatial fields...

Thanks @iranroman. I like @magdalenafuentes 's idea of using `Events` as a parent class - this way you can add the functionality you need without having to reimplement/duplicate all the...

sgtm @magdalenafuentes 👍

Agreed on all points. First question - would it make sense to use the ClipGroup class for representing splits/folds? On the one hand it has relevant features like listing all...

Thanks @magdalenafuentes @pzinemanas. Questions: 1. Yes, this would be great. My question is how do we handle splits vs folds? i.e. some datasets have train/valiate/test/ splits, while others (such as...

1. Yes I agree. Maybe it's fine to use `get_splits()` as the function name and then return the splits/folds following whatever naming convention the dataset creators used? (e.g. some datasets...

I think it would be really great to have it included, since then we have at least 1 speech dataset in our first release and that can really help drive...

Thanks for the detailed report @lostanlen ! Let's start with the jams converter - IIUC the issue is that if you were to execute the example script you have provided...

Thanks @lostanlen , that definitely helps me understand the scenario. It sounds to me (and please correct me if i'm wrong) that there is a tradeoff here: the safest option...