Datasets:
Tasks:
Image-to-Image
Modalities:
Image
Formats:
imagefolder
Languages:
English
Size:
< 1K
ArXiv:
License:
| license: mit | |
| language: | |
| - en | |
| pretty_name: PhyEditBench | |
| size_categories: | |
| - n<1K | |
| task_categories: | |
| - image-to-image | |
| tags: | |
| - image-generation | |
| - image-editing | |
| - benchmark | |
| - physics-aware | |
| <div align="center"> | |
| <h1>PhyEditBench: A Real-World Multi-Stage Benchmark for Physics-Aware Image Editing</h1> | |
| [Shengbin Guo*](https://previsior.github.io/) [Shaokang He*](https://openreview.net/profile?id=~Shaokang_He2) [Chaoyue Meng](https://openreview.net/profile?id=~Chaoyue_Meng1) [Shengpeng Xiao](https://openreview.net/profile?id=~Shengpeng_Xiao2)<br>[Xunzhi Xiang](https://xbxsxp9.github.io/) [Shaofeng Zhang](https://sherrylone.github.io/) [Qi Fan†](https://fanq15.github.io/) | |
| <br> | |
| <sup>* equal contribution † corresponding author</sup> | |
| <p align="center"> | |
| 📖 <a href="https://github.com/Previsior/PhyEditBench"><b>Project Page</b></a>    |    📑 <a href="https://arxiv.org/abs/2606.26551">Arxiv</a>    |    🤗 <a href="https://huggingface.co/datasets/Previsior-PhyEdit/phyedit-video">PhyEditVideo</a> | |
| </p> | |
| </div> | |
| PhyEditBench is a real-world multi-stage benchmark for evaluating physics-aware reasoning in instruction-guided image editing. |