---
tags:
  - text-to-image
  - stable-diffusion
  - lora
  - diffusers
language:
- en
library_name: diffusers
pipeline_tag: text-to-image
base_model: stabilityai/stable-diffusion-2-base
instance_prompt: "Mobile app:"
---

# UI-Diffuser-V2
UI-Diffuser-V2 is fine tuned from "stabilityai/stable-diffusion-2-base" with the [GPSCap dataset](https://paperswithcode.com/dataset/gpscap) for mobile UI generation.

A demo using diffusion model and large language model for UI generation is available at https://github.com/Jl-wei/ai-gen-ui


## Using with Diffusers
```python
import torch
from diffusers import StableDiffusionPipeline, EulerDiscreteScheduler

model_id = "stabilityai/stable-diffusion-2-base"
scheduler = EulerDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler")
pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=scheduler, torch_dtype=torch.float16)

lora_path = "Jl-wei/ui-diffuser-v2"
pipe.load_lora_weights(lora_path)
pipe.to("cuda")

prompt = "Mobile app: health monitoring report"
images = pipe(prompt, num_inference_steps=30, guidance_scale=7.5, height=512, width=288, num_images_per_prompt=10).images

columns = 5
fig = plt.figure(figsize=(20,10))
for i, image in enumerate(images):
    plt.subplot(int(len(images) / columns), columns, i + 1)
    plt.imshow(image)
for ax in fig.axes:
    ax.axis("off")
```

## Citation
If you find our work useful, please cite our paper:
```bibtex
@misc{wei2024aiinspired,
      title={On AI-Inspired UI-Design}, 
      author={Jialiang Wei and Anne-Lise Courbis and Thomas Lambolais and Gérard Dray and Walid Maalej},
      year={2024},
      eprint={2406.13631},
      archivePrefix={arXiv}
}
```
Please note that the code and model can only be used for academic purpose.

### UI-Diffuser-V1
This model, UI-Diffuser-V2, represents the second version of the UI-Diffuser model.

The initial version, UI-Diffuser-V1, was introduced in our paper titled [Boosting GUI Prototyping with Diffusion Models](https://ieeexplore.ieee.org/abstract/document/10260853)