Karine-Huang/T2I-CompBench
[Neurips 2023 & TPAMI] T2I-CompBench (++) for Compositional Text-to-image Generation Evaluation
This project helps researchers and developers assess how well AI models generate images from complex text descriptions. It takes images produced by a text-to-image model and the original descriptive prompts as input. The output is a set of scores indicating how accurately the generated images capture details like object attributes, spatial relationships, and numerical quantities. This is intended for AI researchers, machine learning engineers, and model evaluators who are developing or comparing text-to-image generation systems.
334 stars.
Use this if you need a standardized and comprehensive way to measure the "compositionality" of your text-to-image AI model's outputs.
Not ideal if you are an artist or general user looking for a tool to generate images directly, rather than evaluate an existing model's output.
Stars
334
Forks
18
Language
Python
License
MIT
Category
Last pushed
Dec 24, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/Karine-Huang/T2I-CompBench"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
Vchitect/VBench
[CVPR2024 Highlight] VBench - We Evaluate Video Generation
VectorSpaceLab/OmniGen
OmniGen: Unified Image Generation. https://arxiv.org/pdf/2409.11340
EndlessSora/focal-frequency-loss
[ICCV 2021] Focal Frequency Loss for Image Reconstruction and Synthesis
JIA-Lab-research/DreamOmni2
This project is the official implementation of 'DreamOmni2: Multimodal Instruction-based Editing...
SkyworkAI/UniPic
Open-source SOTA multi-image editing model