Table 1: The top part of the table explains how we will designate the datasets henceforth. For both training and evaluation, we have ground truth denoted as a raw video dataset, the cropped video where the sides were cut off, and the extrapolated dataset. The bottom part of the table indicates the final count of labels categorized either as natural or unnatural in the train and evaluation datasets.

Dataset purpose		Training	Evaluation
Ground truth		Trainraw	Evalraw
Crop 25%		Traincrop	Evalcrop
Extrapolated back		Trainextra	Evalextra
Tag	natural	81	34
	unnatural	225	116
Total		336	150