VanyaBK / visual_ASR_EC

Dataset for the task of visual ASR Error Correction
5 stars 0 forks source link

The vid for videos used #1

Closed jasonppy closed 1 year ago

jasonppy commented 1 year ago

Thanks for the great work and open-source the toolkit.

Is it possible to share the vid of the dataset? I believe that will be much easier for people who want to follow up your work to give a fair comparison. (also the timestamps of the clips if you don't used the entire videos)

VanyaBK commented 1 year ago

Hi, Thank you for your interest in our work. I have uploaded the youtube links which were used. The whole of the video was used in our dataset

jasonppy commented 1 year ago

Thank you for the help. From the file I can see that there are 148k videos, while in the paper it says 127k samples are used (after sentence-bert filtering) so I just want to make sure whether the links are the set after filtering

VanyaBK commented 1 year ago

The links I have posted are before filtering. 127k samples are not the number of videos but the number of parallel sentences pairs(of correct and incorrect sentences) along with images. Hope this answers your question

jasonppy commented 1 year ago

Very helpful, thanks