Views
No views yet
[!WARNING] we do not have a full checkpoint conversion validation, if you encounter pipeline loading failure and unsidered output, please contact me via bili_sakura@zju.edu.cn
| Model | Direction |
|---|---|
horse2zebra | horse -> zebra |
label2cityscape | semantic label -> cityscape |
male2female | male -> female |
mat2sat | map -> satellite |
summer2winter | summer -> winter |
1{model_name}/
2 generator/
3 config.json
4 diffusion_pytorch_model.safetensors1from PIL import Image
2
3from src.pipelines.unsb import UNSBPipeline
4
5pipe = UNSBPipeline.from_pretrained(
6 "/path/to/UNSB-ckpt/horse2zebra",
7 subfolder="generator",
8 device="cuda",
9 scheduler_num_timesteps=5,
10 scheduler_tau=0.01,
11)
12
13source = Image.open("/path/to/source.png").convert("RGB")
14out = pipe(source_image=source, output_type="pil")
15out.images[0].save("unsb_output.png")latest_net_G.pth.src.models.unsb.UNSBGenerator naming in pytorch-image-translation-models:
Dense_time -> dense_timestyle.* parameters removed1@inproceedings{kim2024unpaired,
2 title={Unpaired Image-to-Image Translation via Neural Schroedinger Bridge},
3 author={Kim, Dongjun and others},
4 booktitle={ICLR},
5 year={2024}
6}