v7labs/XMem_ICL

Video Object Segmentation methods are good in-context learners

Flexible visual prompts for in context learning in computer vision Paper

• In-context learning (ICL) for image segmentation.
• Adapts modern Video Object Segmentation (VOS) for visual in-context learning.
• Highly effective for out-of-distribution datasets
• Propose a support set selection technique

Check out our demo

Code

Coming soon!