A training-free pipeline parses a video query with an LLM, tracks candidates with SAM2, then uses coarse motion plus conditional pose reasoning to segment the referred object.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Unleashing Hierarchical Reasoning: An LLM-Driven Framework for Training-Free Referring Video Object Segmentation
A training-free pipeline parses a video query with an LLM, tracks candidates with SAM2, then uses coarse motion plus conditional pose reasoning to segment the referred object.