# SC2GAN: Rethinking Entanglement by Self-correcting Correlated GAN Space

arXiv: https://arxiv.org/abs/2310.06667 · HTML: https://mlsanity.com/paper/2310.06667/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 27 references that are in the map.
- First posted on arXiv 10 Oct 2023.

## Builds on (most cited first, 22 of 22)

- [A Style-Based Generator Architecture for Generative Adversarial Networks](https://mlsanity.com/paper/1812.04948/) (2018)
- [Image-to-Image Translation with Conditional Adversarial Networks](https://mlsanity.com/paper/1611.07004/) (2016)
- [Analyzing and Improving the Image Quality of StyleGAN](https://mlsanity.com/paper/1912.04958/) (2019)
- [Large Scale GAN Training for High Fidelity Natural Image Synthesis](https://mlsanity.com/paper/1809.11096/) (2018)
- [Conditional Generative Adversarial Nets](https://mlsanity.com/paper/1411.1784/) (2014)
- [Alias-Free Generative Adversarial Networks](https://mlsanity.com/paper/2106.12423/) (2021)
- [Image2StyleGAN: How to Embed Images Into the StyleGAN Latent Space?](https://mlsanity.com/paper/1904.03189/) (2019)
- [GANSpace: Discovering Interpretable GAN Controls](https://mlsanity.com/paper/2004.02546/) (2020)
- [StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks](https://mlsanity.com/paper/1612.03242/) (2016)
- [Conditional Image Synthesis With Auxiliary Classifier GANs](https://mlsanity.com/paper/1610.09585/) (2016)
- [Designing an Encoder for StyleGAN Image Manipulation](https://mlsanity.com/paper/2102.02766/) (2021)
- [InterFaceGAN: Interpreting the Disentangled Face Representation Learned by GANs](https://mlsanity.com/paper/2005.09635/) (2020)
- [Image2StyleGAN++: How to Edit the Embedded Images?](https://mlsanity.com/paper/1911.11544/) (2019)
- [StyleSpace Analysis: Disentangled Controls for StyleGAN Image Generation](https://mlsanity.com/paper/2011.12799/) (2020)
- [StyleRig: Rigging StyleGAN for 3D Control over Portrait Images](https://mlsanity.com/paper/2004.00121/) (2020)
- [Disentangled and Controllable Face Image Generation via 3D Imitative-Contrastive Learning](https://mlsanity.com/paper/2004.11660/) (2020)
- [High-Fidelity GAN Inversion for Image Attribute Editing](https://mlsanity.com/paper/2109.06590/) (2021)
- [Controllable Text-to-Image Generation](https://mlsanity.com/paper/1909.07083/) (2019)
- [Seeing What a GAN Cannot Generate](https://mlsanity.com/paper/1910.11626/) (2019)
- [Improved StyleGAN Embedding: Where are the Good Latents?](https://mlsanity.com/paper/2012.09036/) (2020)
- [Provable Domain Generalization via Invariant-Feature Subspace Recovery](https://mlsanity.com/paper/2201.12919/) (2022)
- [Exploring Gradient-based Multi-directional Controls in GANs](https://mlsanity.com/paper/2209.00698/) (2022)

## Abstract (arXiv)

Generative Adversarial Networks (GANs) can synthesize realistic images, with the learned latent space shown to encode rich semantic information with various interpretable directions. However, due to the unstructured nature of the learned latent space, it inherits the bias from the training data where specific groups of visual attributes that are not causally related tend to appear together, a phenomenon also known as spurious correlations, e.g., age and eyeglasses or women and lipsticks. Consequently, the learned distribution often lacks the proper modelling of the missing examples. The interpolation following editing directions for one attribute could result in entangled changes with other attributes. To address this problem, previous works typically adjust the learned directions to minimize the changes in other attributes, yet they still fail on strongly correlated features. In this work, we study the entanglement issue in both the training data and the learned latent space for the StyleGAN2-FFHQ model. We propose a novel framework SC2GAN that achieves disentanglement by re-projecting low-density latent code samples in the original latent space and correcting the editing directions based on both the high-density and low-density regions. By leveraging the original meaningful directions and semantic region-specific layers, our framework interpolates the original latent codes to generate images with attribute combination that appears infrequently, then inverts these samples back to the original latent space. We apply our framework to pre-existing methods that learn meaningful latent directions and showcase its strong capability to disentangle the attributes with small amounts of low-density region samples added.