Traditional self-supervised learning requires convolutional neural networks (CNNs) using external pretext tasks (i. e. image- or video-based tasks) to encode high-level semantic visual representations. https://www.duospiritalis.com/