CIFAR10 and CIFAR100

The dataset used in the paper is not explicitly described, but it is mentioned that the authors conducted experiments on various vision tasks, including image classification, object detection, human-object interaction detection, and human body shape recovery.