You can deploy this model and start earning money today!

mediasynthesismuseum/
pix2pix

Name: pix2pix
Availability: InStock

Views

No views yet

Pix2Pix pretrained generators (PyTorch)

The original Pix2Pix (Isola et al., CVPR 2017, Image-to-Image Translation with Conditional Adversarial Networks) pretrained U-Net generators, hosted for mediasynthesismuseum/pix2pix. These are the official berkeley PyTorch weights (junyanz/pytorch-CycleGAN-and-pix2pix).

File	Task	Input
`edges2shoes.pth`	edges → shoe photo	draw an outline
`edges2handbags.pth`	edges → handbag photo	draw an outline
`facades_label2photo.pth`	label map → building facade	draw colored regions
`day2night.pth`	daytime → nighttime photo	upload a photo
`map2sat.pth`	map tile → satellite image	upload a map

All are unet_256 generators: input_nc=output_nc=3, ngf=64, BatchNorm.

Preprocessing / usage

Resize to 256×256, normalize to [-1, 1] ((x/127.5) - 1), RGB; de-normalize the output the same way. The UnetGenerator definition is included in this repo (pix2pix_net.py).

python

import torch
from huggingface_hub import hf_hub_download
from pix2pix_net import build_generator

G = build_generator()
sd = torch.load(hf_hub_download("mediasynthesismuseum/pix2pix", "edges2shoes.pth"),
                map_location="cpu", weights_only=False)
G.load_state_dict(sd); G.eval()

Credit: Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, Alexei A. Efros (pix2pix), and the pytorch-CycleGAN-and-pix2pix authors.