What are the different data augmentation techniques followed in the paper?

[ajay9022]

In the paper, accuracy for different data-augmentation techniques have been such as

**1. Random-clipping

2. Multiscale-cropping**

I didn't understand what they mean. I read something on Random-clipping but it wasn't enough to understand.

Can you put some light on these two?

[gulvarol]

Random clipping is selecting a random starting frame from the sequence, which is quite standard nowadays. The implementation is pretty clear: https://github.com/gulvarol/ltc/blob/master/donkey.lua#L42

Multiscale cropping is a simple augmentation as well: https://github.com/gulvarol/ltc/blob/master/donkey.lua#L213

If the explanations in the paper are not sufficient to understand, you can also check the references we give ([6, 23]).