Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• SonoWorld is the first framework to tackle the Image2AVScene task of generating a 3D audio-visual scene from a single image.method 95%
• The SonoWorld pipeline outperforms prior spatial audio generation methods by reducing Direction of Arrival (DOA) error by 47%.result 90%
• The SonoWorld pipeline uses language-guided sound anchor placement to align audio with scene semantics.method 85%
• The method is evaluated on a newly curated real-world dataset called SonoScene360.method 90%
• SonoWorld demonstrates applications beyond rendering, including one-shot acoustic learning and audio-visual spatial source separation.technical 80%
• The pipeline's panorama-to-3D reconstruction component can utilize multiple existing methods, with Marble model noted for offering the best visual rendering quality.technical 85%
• The method significantly outperforms baselines on semantic audio quality metrics (D-CLAP), improving scores by more than 39% over the state-of-the-art video-to-audio baseline.result 90%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card