Paper

The Target Is a Predicate: Predicate-State Grounding for Semantic-Level Video Reasoning Segmentation · ByteTrail