Iran R. Roman

I think what we need is a `SpatialEvents` class that accounts for `azimuth`, `elevation`, and `distance` of events. This would be a simple `Events` class spinoff. ``` class SpatialEvents(Annotation): """Events...

Hello again, Thanks for your suggestions. I've been making significant progress on the loader. I have a question and want to hear your thoughts though. The soundata `Events` class uses...

Hello @magdalenafuentes. Thanks for your thoughts. This dataset has timestamps for each 100ms of data, so it does not just have a `[start_time, end_time]` for a given event. I have...

> IIUC the events are annotated as time series of active/silent with a 100ms hop size. Is that correct? That is correct. In the original dataset annotations, the following are...

@magdalenafuentes it is true that out of the three position (azimuth, elevation, distance) only one is required. Some datasets only explore one or two of these dimensions for localization. _think...

I studied the `seld-dcase2021` code to understand how they process the annotations. They convert the csv values into a dictionary where the `time-frames` are the keys, and the values are...

I agree! These two points would be very useful for microphone array datasets in general. We can include them. I'll be happy to write them. What does @justinsalamon (and others)...

I agree that expanding the existing `Event` type should be good enough. Is there a way for us to make the Event type flexible enough to have it include ANY...

I agree. Just to confirm, you are saying that each "spatial" dataset loader will independently define this `SpatialEvents` class inheriting from `Events`, and will adjust it according to what its...

Given everything we have discussed, I believe that `SoundEvents` should have the following arguments: ``` Attributes: intervals (np.ndarray): (n x 2) array of intervals (as floats) in seconds in the...