Open xiaoxue1117 opened 1 year ago
Thanks for following our work! The data for VAD training of IMU encoder is separated from the ALT training. The format of data is a bit different. Let me add the data to the Zenodo now ... Before that, you can use a checkpoint of VAD model instead so that no need to train yourself. I'll upload it to the ImuVAD folder.
(They are in 5-second segment format instead of utterance-level, with some additional columns)
(They are in 5-second segment format instead of utterance-level, with some additional columns)
Thank you for your reply! If you can also update the fusion module for multi-modal training, that would be great!
Cannot access key "label" in the row 334 for main.py under ImuVAD directory.