Large-scale text-to-image diffusion models have made amazing advances. However, the status quo is to use text input alone, which can impede controllability. In this work, we propose GLIGEN, Grounded-Language-to-Image Generation, a novel approach that builds upon and extends the functionality of existing pre-trained text-to-image diffusion models by enabling them to also be conditioned on grounding inputs. To preserve the vast concept knowledge of the pre-trained model, we freeze all of its weights and inject the grounding information into new trainable layers via a gated mechanism. Our model achieves open-world grounded text2img generation with caption and bounding box condition inputs, and the grounding ability generalizes well to novel spatial configurations and concepts. GLIGEN's zero-shot performance on COCO and LVIS outperforms that of existing supervised layout-to-image baselines by a large margin.
@article{arxiv.2301.07093,
title = {GLIGEN: Open-Set Grounded Text-to-Image Generation},
author = {Yuheng Li and Haotian Liu and Qingyang Wu and Fangzhou Mu and Jianwei Yang and Jianfeng Gao and Chunyuan Li and Yong Jae Lee},
journal= {arXiv preprint arXiv:2301.07093},
year = {2023}
}