A new large audio-image-text dataset and a FiLM-based audio-to-image model are presented, with strong in-house metrics but evaluation caveats on the external set.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Towards Expressive and Faithful Audio-to-Image Generation: A Unified Multimodal Dataset and Synthesis Framework
A new large audio-image-text dataset and a FiLM-based audio-to-image model are presented, with strong in-house metrics but evaluation caveats on the external set.