THU-LYJ-Lab/T3Bench
A benchmark suite for evaluating text-to-3D generation models using automatic quality and alignment metrics driven by LLMs.
Not currently ranked — collecting fresh signals.
star history
T3Bench provides 300 text prompts across three complexity levels to assess text-to-3D generation methods. It proposes two automatic evaluation metrics: a quality metric combining multi-view text-image scores with regional convolution to detect quality and view inconsistency, and an alignment metric using multi-view captioning with LLM evaluation to measure text-3D consistency. The benchmark integrates with ThreeStudio for generating 3D content from prompts and provides a standardized framework for comparing progress in the text-to-3D domain.
Frequently asked
- What is THU-LYJ-Lab/T3Bench?
- A benchmark suite for evaluating text-to-3D generation models using automatic quality and alignment metrics driven by LLMs.
- Is T3Bench open source?
- Yes — THU-LYJ-Lab/T3Bench is an open-source project tracked on heatdrop.
- What language is T3Bench written in?
- THU-LYJ-Lab/T3Bench is primarily written in Python.
- How popular is T3Bench?
- THU-LYJ-Lab/T3Bench has 1.1k stars on GitHub.
- Where can I find T3Bench?
- THU-LYJ-Lab/T3Bench is on GitHub at https://github.com/THU-LYJ-Lab/T3Bench.