high-resolution paired image-to-image translation with GANs
Official PyTorch implementation of pix2pixHD, a conditional GAN for high-resolution paired image-to-image translation, typically from semantic label maps to photorealistic street scenes or faces.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Unsupervised multimodal image-to-image translation; choose when you have unpaired domains and want diverse outputs rather than paired label-map supervision.
Multimodal Unsupervised Image-to-Image Translation
Unsupervised unpaired image translation without cycle consistency; choose for fast and lightweight training on unpaired datasets.
Contrastive unpaired image-to-image translation, faster and lighter training than cyclegan (ECCV 2020, in PyTorch)
Unsupervised unpaired image translation with shared latent space; earlier model useful for simple unpaired-domain translation.
Unsupervised Image-to-Image Translation
One-step diffusion-based image-to-image translation; choose when you can use pre-trained Stable Diffusion and want fast inference without training a GAN.
One-step image-to-image with Stable Diffusion turbo: sketch2image, day2night, and more
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
Official lower-resolution pix2pix/CycleGAN codebase; choose for standard 256px paired or unpaired translation instead of HD semantic synthesis.
Image-to-Image Translation in PyTorch
TensorFlow port of the original low-resolution pix2pix; choose only when TensorFlow is required and 256px output is sufficient.
Tensorflow port of Image-to-Image Translation with Conditional Adversarial Nets https://phillipi.github.io/pix2pix/
These projects were analysed and named pix2pixHD among their alternatives. The relationship is not symmetric — how pix2pixHD rates them is a separate judgement, made when pix2pixHD is analysed in its own right.
calls pix2pixHD “Choose for high-resolution (2048x1024) paired translation with more detailed output than standard pix2pix.”
Image-to-Image Translation in PyTorch
calls pix2pixHD “Choose for high-resolution (2048x1024) conditional image synthesis with a dedicated PyTorch repo, rather than PaddleGAN's broader library.”
PaddlePaddle GAN library, including lots of interesting applications like First-Order motion transfer, Wav2Lip, picture repair, image editing, photo2cartoon, image style transfer, GPEN, and so on.