I have a set of video frames of human faces. I am using trying use GNN for predictions. Can I get insights into how to create the Dataset.
Structure of root:
/root/ |--> sub1/ |--> video1/ |--> frame_xy....
when processing, what would be the best way to save to the processing_dir for each subject
[sub1-video1``: [Data(frame1, adj1, y1), Data(frame2, adj2, y2),....], sub1_video2: [Data(frame1, adj1, y1), Data(frame2, adj2, y2`),....]
so on and so forth. Is there a better way to represent it? Can I be pointed towards an example dataset.
Thanks in advance, Regards
I have looked at the pytorch-geometric docs: https://pytorch-geometric.readthedocs.io/en/latest/modules/datasets.html
Also seen some example on github, here is one: https://github.com/deepfindr/gnn-project/blob/main/dataset.py