Training/evaluating on all sequential frames

[jamalknight]

This model skips redundant frames (where the subject is not moving) and therefore does not process all frames. If I wanted to test a sequence of all frames (not skipping any) could I:

1. use this model trained on skipped frames (as it currently is) and evaluate it on a sequence where no frames are skipped

or

2. would I have to retrain using data that does not skip frames?

Ultimately I would like to have smooth sequential motion without frames skipped and wondering what the best approach would be.

Thanks

[shrubb]

No, you don't have to retrain. The model will work already. It doesn't "skip redundant frames", it's just that the dataset was preprocessed this way. That doesn't matter.

[jamalknight]

Great, thanks