LayoutBench-COCO

The evaluation code for LayoutBench-COCO, a new diagnostic benchmark for layout-guided image generation for real objects, as described in the paper:

Diagnostic Benchmark and Iterative Inpainting for Layout-Guided Image Generation (CVPR 2024 Workshop)

Jaemin Cho, Linjie Li, Zhengyuan Yang, Zhe Gan, Lijuan Wang, Mohit Bansal

[Project Page] [Paper] [Dataset (HF Hub)]

Please see LayoutBench Github for CLEVR-based LayoutBench dataset.

Examples of LayoutBench-COCO input layouts in 4 skill splits.

Usage

Step 1. Download LayoutBench-COCO and Generate Images from Layouts

Please see ./image_generation/README.md for instructions on downloading LayoutBench-COCO and generating images from layouts.

Example generated images

Step 2. Run Evaluation on LayoutBench-COCO

Please see ./yolov7/README.md for instructions on running evaluation on LayoutBench-COCO.

Example evaluation results

Citation

If you find our project useful in your research, please cite the following paper:

@inproceedings{Cho2024LayoutBench,
  author    = {Jaemin Cho and Linjie Li and Zhengyuan Yang and Zhe Gan and Lijuan Wang and Mohit Bansal},
  title     = {Diagnostic Benchmark and Iterative Inpainting for Layout-Guided Image Generation},
  booktitle = {The First Workshop on the Evaluation of Generative Foundation Models},
  year      = {2024},
}