PromptonomyViT: Multi-Task Prompt Learning Improves Video Transformers using Synthetic Scene DataRoei HerzigOfir Abramovichet al.2024WACV 2024