Random splitting strategy is a common approach for training, testing, and validating object detection algorithms based on deep learning. Is common for datasets to have images extracted from video ...