Training input format

[shihlun1208]

I have downloaded the jester datasets, it contains RGB video frame and label csv file.
The model seems can be trained in RGB or RGBD format by modifying the modality to RGB or depth.
Could you provide the depth image training input?
For example:
In Jester RGB format
Folder 34870 - video sequence of Drumming Fingers into frames
train.csv -
34870;Drumming Fingers
56557;Sliding Two Fingers Right
.....