Direct label-space diffusion, conditioned on image and text, achieves state-of-the-art 69.7 Average Recall on Panoptic Narrative Grounding.
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
-
GS: Generative Segmentation via Label Diffusion
Direct label-space diffusion, conditioned on image and text, achieves state-of-the-art 69.7 Average Recall on Panoptic Narrative Grounding.