Dynamic Sampling Of Video To Imagery For Deep Learning