OverLayBench: A Benchmark for Layout-to-Image Generation with Dense Overlaps
Bingnan Li, Chen-Yu Wang, Haiyang Xu, Xiang Zhang, Ethan J. Armand, Divyansh Srivastava, Xiaojun Shan, Zeyuan Chen
Abstract
Despite steady progress in layout-to-image generation, current methods still struggle with layouts containing significant overlap between bounding boxes. We identify two primary challenges: (1) large overlapping regions and (2) overlapping instances with minimal semantic distinction. Through both qualitative examples and quantitative analysis, we demonstrate how these factors degrade generation quality. To systematically assess this issue, we introduce OverLayScore, a novel metric that quantifies the complexity of overlapping bounding boxes. Our analysis reveals that existing benchmarks are biased toward simpler cases with low OverLayScore values, limiting their effectiveness in evaluating models under more challenging conditions. To reduce this gap, we present OverLayBench, a new benchmark featuring balanced OverLayScore distributions and high-quality annotations. As an initial step toward improved performance on complex overlaps, we also propose CreatiLayout-AM, a model trained on a curated amodal mask dataset. Together, our contributions establish a foundation for more robust layout-to-image generation under realistic and challenging scenarios.
BibTeX
@inproceedings{
li2025overlaybench,
title={OverLayBench: A Benchmark for Layout-to-Image Generation with Dense Overlaps},
author={Bingnan Li and Chen-Yu Wang and Haiyang Xu and Xiang Zhang and Ethan J. Armand and Divyansh Srivastava and Xiaojun Shan and Zeyuan Chen and Jianwen Xie and Zhuowen Tu},
booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
year={2025},
url={https://openreview.net/forum?id=FDi6Mbl0xD}
}