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

BiliSakura/pix2pix-sar2rgb-ckpt

Checkpoint-style packaging of yuulind/pix2pix-sar2rgb, aligned with examples/community/sar2optical in pytorch-image-translation-models.

Variants

Variant	Source epoch
epoch-180	180
epoch-265	265
epoch-295	295

Repository layout

pix2pix-sar2rgb-ckpt/
  epoch-180/
    generator/
      config.json
      diffusion_pytorch_model.safetensors
    discriminator/
      config.json
      diffusion_pytorch_model.safetensors
  epoch-265/
    generator/
      config.json
      diffusion_pytorch_model.safetensors
    discriminator/
      config.json
      diffusion_pytorch_model.safetensors
  epoch-295/
    generator/
      config.json
      diffusion_pytorch_model.safetensors
    discriminator/
      config.json
      diffusion_pytorch_model.safetensors

Usage (Generator Inference)

from PIL import Image

from examples.community.sar2optical.pipeline import SAR2OpticalPipeline

pipe = SAR2OpticalPipeline.from_pretrained(
    "/path/to/pix2pix-sar2rgb-ckpt/epoch-295",
    subfolder="generator",
    device="cuda",
)

sar = Image.open("/path/to/sar_input.png").convert("RGB")
out = pipe(source_image=sar, output_type="pil")
out.images[0].save("sar2opt_output.png")

Notes

• Converted from upstream .pth generator/discriminator checkpoints.
• All three epochs load strictly with SAR2OpticalGenerator and SAR2OpticalDiscriminator.

Citation

@inproceedings{isola2017pix2pix,
  title={Image-to-Image Translation with Conditional Adversarial Networks},
  author={Isola, Phillip and Zhu, Jun-Yan and Zhou, Tinghui and Efros, Alexei A.},
  booktitle={CVPR},
  year={2017}
}