ViT + MAE for UDA on Sentinel-1/2 (SAR/optical) land-cover classification with CORAL & DANN. PyTorch.

Masked autoencoders — implementation and evaluation for vision tasks using Vision Transformers (ViTs)

PySlowFast, the official video understanding framework from Facebook AI Research (FAIR), to train, evaluate, and reproduce state-of-the-art video models on the UCF24 action detection dataset. It supports customizable training pipelines, model fine-tuning, and evaluation for video-based action recognition and spatio-temporal localization tasks.