A framework that generates image-level object concepts with a vision-language model before region segmentation improves open-vocabulary segmentation on multiple benchmarks.
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
-
What You Perceive Is What You Conceive: A Cognition-Inspired Framework for Open Vocabulary Image Segmentation
A framework that generates image-level object concepts with a vision-language model before region segmentation improves open-vocabulary segmentation on multiple benchmarks.