Luffuly's picture
update readme
7487caa
|
raw
history blame
1.24 kB
metadata
license: mit

Unique3d-MVImage-Diffuser Model Card

Example

Note the input image is suppose to be white background.

import torch 
import numpy as np 
from PIL import Image 
from pipeline_img2mvimg import StableDiffusionImage2MVCustomPipeline


pipe = StableDiffusionImage2MVCustomPipeline.from_pretrained( 
    "Luffuly/unique3d-mv-variation-diffuser",  
    torch_dtype=torch.float16, 
    trust_remote_code=True 
).to("cuda")


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

forward_args = dict(
    width=256,
    height=256,
    width_cond=256,
    height_cond=256,  
    num_images_per_prompt=4, 
    num_inference_steps=50,  
    guidance_scale=1.5,  
) 

out = pipe(image, **forward_args).images
rgb_np = np.hstack([np.array(img) for img in out])
Image.fromarray(rgb_np).save(f"mvimg.png")

Image-to-Normal

please check:

Citation

@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}
}