VIDEO OBJECT SEGMENTATION WITH WORLD KNOWLEDGE

Summary from the TLDR AI newsletter

Reasoning Video Object Segmentation (ReasonVOS) is a new task that uses implicit text queries to generate segmentation masks. It requires complex reasoning and world knowledge.