Raw audio clip data and their pose data.

[chuangg]

Hi Authors:

Thanks for the great work. I would like to explore the different sound representations. Could you please provide the clip information used in your training and testing data? I don't find a solution on how to align the raw audio data with the pre-processed pose data.

Thanks!

[shlizee]

Hello chuangg,

You can find the list of videos (audio and frames) included in the training dataset in the data.zip file. The filename is list_of_recital_videos.txt.

You can find instruction on how to download the file in the readme of the data folder: https://github.com/facebookresearch/Audio2BodyDynamics/tree/master/data

[chuangg]

Thanks for your reply! I think you might misinterpret my question. I know that you provide the raw videos, but still, have no ideas on how to align the raw video with the provided post-processed body pose.

It will be very nice if you can provide the raw audio file. Then we can have a detailed study on the sound representations.

Thanks

[shlizee]

This is just a sample of videos. We cannot supply their raw data or references as this is third party data, but you have several alternatives:

1. Re-train the network on your raw audio and use MFCC / other features than MFCC for training
2. Run MFCC features detector on the referred sampled videos and align the MFCC features
3. Implement the pre-processing/filtering we describe in the paper

[chuangg]

Thanks again for your reply. I fully understand that you can not provide the raw data and third-party code. But could you please provide the time interval (e.g. starting and ending time) for each sound/pose clips used for training and testing? Then we can easily use this information to extract the audio clips and align with the pre-processed pose.

Thanks so much!

[chuangg]

Hi professor Eli:

Could you please let us know if you can release the time interval (starting and ending time) for each clip? I saw that there are also many other researchers requiring the same data. It will be very nice to let us know first if you can provide it or not.

Thanks Chuang