Hello, thanks for your work and for sharing the code, I'd like to know whether I can use it to predict a single video without re-training.