Multimodal (Audio, Facial and Gesture) based Emotion Recognition challenge Forum

Go back to competition Back to thread list Post in this thread

> The files in "phase_2/motion capture" don't follow the format in the readme file

I noticed that the file names and structure don't follow the convention in the readme file, although the target information seems to be available in the filenames and mocap frames are already segmented.

Posted by: ronnypetson @ Nov. 18, 2019, 4:01 p.m.

Thank you for reporting this, however the description provided is not sufficient to locate the issue, can you please provide an example of the case you are describing.

Posted by: dam @ Nov. 18, 2019, 5:54 p.m.

The readme says:

"File content format:
[...]
Line 5 - kinect section header
Line 6 - source motion capture file name
Line 7 - frame index of the beginning of the expression in the source motion capture file
Line 8 - frame index of the end of the expression in the source motion capture file
"

But if you check the files in "phase_2/descriptors/", for instance frames_F2_An1.txt, the line 6 will contain something like "20170902_085103_00.txt", but there is no such file. Instead, in the folder "motion capture" there are files like M8_Su5.txt, that are already segmented versions of the kinect samples.

Posted by: ronnypetson @ Nov. 19, 2019, 1:22 p.m.

I am sorry for this issue. We will fix it today. Of course the deadline will be extended.

Posted by: dam @ Dec. 12, 2019, 10:26 p.m.

The issue is fixed. Once again sorry for inconvenience.

Posted by: dam @ Dec. 12, 2019, 11:23 p.m.

There is also a typo in descriptor frames_M1_Sa5.txt. On line 2, there is a reference to video "M1_927_0160_01.MP4", which does not exist. We guess that it should have been "927_0160_01.MP4" instead.

Posted by: DataLabUPB @ Dec. 23, 2019, 9:39 a.m.

Thank you, you are right. We will correct that mistake today.

Posted by: dam @ Jan. 5, 2020, 7:13 p.m.
Post in this thread