This repository serves as a comprehensive collection of resources for GAN inversion, supplementing a TPAMI 2022 survey paper. It targets researchers and practitioners in generative AI, offering a curated list of papers, implementations, and pretrained models for GAN inversion and latent space editing across 2D and 3D domains, as well as diffusion models.

How It Works

The repository is structured as a curated bibliography, categorizing research papers and associated code/models based on their contribution to GAN inversion, diffusion inversion, latent space editing, and various applications. It provides links to official project pages, code repositories, and PDFs for each entry, facilitating easy access to the underlying research.

Quick Start & Requirements

This repository is a collection of links and does not have a direct installation or execution command. Users are expected to follow the links provided for individual papers and their associated codebases, which may have their own specific requirements (e.g., Python, PyTorch, TensorFlow, specific hardware).

Highlighted Details

• Extensive coverage of GAN inversion methods for both 2D and 3D generative models.
• Includes a dedicated section for diffusion model inversion and editing techniques.
• Lists numerous applications of GAN inversion, ranging from image restoration to 3D reconstruction and face recognition.
• Provides links to code implementations and project pages for most listed papers.

Maintenance & Community

The repository is maintained by Weihao Xia and collaborators, with an open invitation for pull requests to suggest missing works or improvements.

Licensing & Compatibility

The repository itself is a collection of links and does not impose a specific license. The licensing of individual code repositories and models linked within will vary and must be checked separately.

Limitations & Caveats

This is a survey and resource collection, not a runnable library. Users must independently locate and set up the code for each specific method they wish to use. The rapid pace of research means some links or code might become outdated.