PAT3D: A Multimodal Benchmark for Physics-Augmented Text-to-3D Scene Generation

A multimodal benchmark dataset associated with the PAT3D research project, created by author guyingl and last updated on 2026-05-18. It contains text prompts, reference images, and generated 3D scene outputs for evaluating text-to-3D generation methods. The dataset emphasizes physically plausible, intersection-free, and simulation-ready scenes.

Use Cases

• Benchmarking text-to-3D scene generation models based on the provided prompts and reference outputs.
• Evaluating the physical plausibility and intersection-free properties of generated 3D scenes.
• Training or fine-tuning models for simulation-ready 3D asset creation using the multimodal prompts and outputs.

Strengths

• Dataset is explicitly designed as a benchmark for evaluating text-to-3D generation, providing a clear use case.
• Focuses on physically plausible and simulation-ready scenes, a specific and relevant quality for 3D generation.

Limitations

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

Provenance

Source

huggingface

Collection Method

Likely generated as part of the associated research project 'PAT3D: Physics-Augmented Text-to-3D Scene Generation'.

Freshness

Last updated 2026-05-18 19:37:49; freshness should be verified.