Segmented segmented videos in CMU_MOSEI are corrupted.

[e-remington-lee]

Hello!

I saw that some videos in the CMI_MOSEI segmented section are corrupted. The videos only have audio (no sound) and open-cv fails to open the video and extract frames. One example of this is the "_0efYOjQYRc_0" which is the first video in the segmented section. a few others are "_4K620KW_Is_7.mp4, _4PNh8dIILI_0.mp4"

I tried redownloading but it seems I get the same result each time. I can come up with a script to resegment the videos based on the time tables I have in the transcripts folder, but it would be super helpful if there might be a solution to this problem!

Thanks a lot for your work

[A2Zadeh]

Hi @e-remington-lee,

This seems to be an issue on our side. I am not sure what happened there, we do use RAID machines to store our datasets. I will look more into this.

In the meantime, you can re-extract the segments using ffmpeg. I am currently a bit busy releasing the CMU-MOSEAS dataset, but can look into this a bit further at the end of October.

[cj00719]

@e-remington-lee Thank you for the question. I recently downloaded the dataset and observed the same issues. Please kindly help to share your segmented version for the videos. Thank you

In addition, can you assist in understanding the labeling of the dataset?

Thank you very much