WFLW

New self-supervised learning method enhances representation for symmetric data

Researchers have introduced Mirror-Fusion-Augmented Self-Supervised Learning (MFASSL), a framework designed to improve representation learning, particularly for data with bilateral symmetry. Unlike standard methods that…