Micro-expression recognition with supervised contrastive learning

. The results on CASME II and SAMM datasets show that the approach is effective and competitive compared with the state-of-the-art methods both on three-classes and five-classes evaluations.