3DIS-FLUX: simple and efficient multi-instance generation with DiT rendering
Paper • 2501.05131 • Published • 37
import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("sanaka87/3DIS", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]News
Our work is accepted by ICLR 2025 (Spotlight) !
Usage
This is the space for the pretrained-weights of 3DIS. The inference code can be found in our GitHub Repository.
Citation
If you find this repository useful, please use the following BibTeX entry for citation.
@article{zhou20243dis,
title={3dis: Depth-driven decoupled instance synthesis for text-to-image generation},
author={Zhou, Dewei and Xie, Ji and Yang, Zongxin and Yang, Yi},
journal={arXiv preprint arXiv:2410.12669},
year={2024}
}
@article{zhou20253disflux,
title={3DIS-FLUX: simple and efficient multi-instance generation with DiT rendering},
author={Zhou, Dewei and Xie, Ji and Yang, Zongxin and Yang, Yi},
journal={arXiv preprint arXiv:2501.05131},
year={2025}
}
- Downloads last month
- 55
Model tree for sanaka87/3DIS
Base model
black-forest-labs/FLUX.1-dev