Why does this Pytorch video tutorial (and others) only work with uniform clip sample lengths?
KevinJMcEntee opened this issue · comments
KevinJMcEntee commented
I have a training set of variable length videos where the anomalies that lead to accurate classification are sparse. If I only train on a short fixed clip from each video then there is a high likelihood the selected clip will not contain the pertinent anomaly needed for classification.