Deep learning has revolutionized various fields by setting new benchmarks that were previously unattainable. Yet, these advancements come at a cost: the necessity for vast amounts of data, which is often expensive and labor-intensive to collect, especially if it requires labeling or cleaning. To address this challenge, researchers have pioneered the field of Self-Supervised Learning (SSL), exploring innovative ways to train deep learning models without the need for labeled data or extensive data cleaning. In this presentation, I will first provide an overview of existing SSL paradigms, then delve into cutting-edge methods, focusing on harnessing unlabeled videos to learn powerful image representations. This approach exemplifies the shift towards more efficient and accessible deep learning methodologies.