Hi there. I am quite new to PyTorch. I just finished the getting started tutorial section. I am searching for some examples for loading video datasets. I looked at https://video-dataset-loading-pytorch.readthedocs.io/en/latest/ but what I understood is that as I am working for human recognition from the way a person walks, I don’t need segments of frames from the sequences rather I need the whole. So, I am kinda stuck because the videos have varying lengths and the videos have different resolutions. How can I efficiently load videos or frame sequences of varying lengths and then split it into train validation test sets?
My dataset looks like this:
dataset_folder
------person_1
|----------- walking sequence 1
|------------- img_0001.jpg
.
.
.
|------------- img_0039.jpg
|----------- walking sequence 2
|------------- img_0001.jpg
.
.
.
|------------- img_0045.jpg
------person_2
|----------- walking sequence 1
|------------- img_0001.jpg
.
.
.
|------------- img_0045.jpg
------person_3
|----------- walking sequence 2
|------------- img_0001.jpg
.
.
.
|------------- img_0038.jpg
|----------- walking sequence 2
|------------- img_0001.jpg
.
.
.
|------------- img_0027.jpg