README.md

---
license: mit
---
# Unique3d-Normal-Diffuser Model Card 

[🌟GitHub](https://github.com/TingtingLiao/unique3d_diffuser) | [🦸 Project Page](https://wukailu.github.io/Unique3D/) | [🔋MVImage Diffuser](https://huggingface.co/Luffuly/unique3d-mvimage-diffuser) 

![mv-normal](https://github.com/user-attachments/assets/de91a83b-a14f-4878-a950-4d5cba786f69)

## Example  
Note the input image is suppose to be **white background**. 

![mv-normal](https://github.com/user-attachments/assets/f0b56d70-d1fb-4f18-a205-f41f85ec72d7)


```bash 
import torch 
import numpy as np 
from PIL import Image 
from pipeline import Unique3dDiffusionPipeline 

# opts 
seed = -1    
generator = torch.Generator(device='cuda').manual_seed(-1)
forward_args = dict(
    width=512,
    height=512, 
    width_cond=512,
    height_cond=512, 
    generator=generator,
    guidance_scale=1.5,   
    num_inference_steps=30, 
    num_images_per_prompt=1, 
)  

# load 
pipe = Unique3dDiffusionPipeline.from_pretrained( 
    "Luffuly/unique3d-normal-diffuser", 
    torch_dtype=torch.bfloat16, 
    trust_remote_code=True,  
).to("cuda")  

# load image 
image = Image.open('image.png').convert("RGB") 

# forward 
out = pipe(image, **forward_args).images 
out[0].save(f"out.png")
```

## Citation
```bash 
@misc{wu2024unique3d,
      title={Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image}, 
      author={Kailu Wu and Fangfu Liu and Zhihan Cai and Runjie Yan and Hanyang Wang and Yating Hu and Yueqi Duan and Kaisheng Ma},
      year={2024},
      eprint={2405.20343},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}
```