Egocentric video understanding, video-language models, action recognition, robustness, and parameter fine-tuning.
No public activity