MARVEL-40M+: Multi-Level Visual Elaboration for Text-to-3D Generation

MARVEL-40M+ is a dataset for high-fidelity text-to-3D content creation, introduced in a CVPR 2025 paper. It appears to be a multi-level visual elaboration dataset, likely containing 3D assets and associated data. The dataset was released by authors including Sankalp Sinha and Mohammad Sadil Khan.

Use Cases

• Training text-to-3D generative models based on multi-level visual elaboration data.
• Benchmarking 3D content generation systems using the described high-fidelity assets.
• Researching multi-modal learning for 3D object creation based on the dataset's stated purpose.

Strengths

• Associated with a CVPR 2025 publication, indicating peer-reviewed research context.
• Designed for a specific, high-demand task: high-fidelity text-to-3D content creation.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Freshness

Last updated 2025-11-10 15:04:19; freshness should be verified.