EPIC_SOUNDS data problems

[WannaSir]

i have got the HDF5 file of EPIC_SOUNDS, but how to split it into train test and validation?

[JacobChalk]

Hi,

Information on the EPIC-Sounds splits can be found here. We also have annotations compatible with TIM here.

To obtain the samples for a specific video ID in python, you need to do samples = audio_dataset[video_id][()] where audio_dataset is the read in HDF5 file. See the baseline code here for more details.