# Condition Matters in Full-head 3D GANs

arXiv: https://arxiv.org/abs/2602.07198 · HTML: https://mlsanity.com/paper/2602.07198/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 29 of its 36 references that are in the map.
- First posted on arXiv 6 Feb 2026.

## Builds on (most cited first, 29 of 29)

- [A Style-Based Generator Architecture for Generative Adversarial Networks](https://mlsanity.com/paper/1812.04948/) (2018)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Deep Learning Face Attributes in the Wild](https://mlsanity.com/paper/1411.7766/) (2014)
- [Efficient Geometry-aware 3D Generative Adversarial Networks](https://mlsanity.com/paper/2112.07945/) (2021)
- [FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space](https://mlsanity.com/paper/2506.15742/) (2025)
- [pi-GAN: Periodic Implicit Generative Adversarial Networks for 3D-Aware Image Synthesis](https://mlsanity.com/paper/2012.00926/) (2020)
- [GRAF: Generative Radiance Fields for 3D-Aware Image Synthesis](https://mlsanity.com/paper/2007.02442/) (2020)
- [Pivotal Tuning for Latent-based Editing of Real Images](https://mlsanity.com/paper/2106.05744/) (2021)
- [StyleSDF: High-Resolution 3D-Consistent Image and Geometry Generation](https://mlsanity.com/paper/2112.11427/) (2021)
- [GRAM: Generative Radiance Manifolds for 3D-Aware Image Generation](https://mlsanity.com/paper/2112.08867/) (2021)
- [BlockGAN: Learning 3D Object-aware Scene Representations from Unlabelled Images](https://mlsanity.com/paper/2002.08988/) (2020)
- [3D Shape Induction from 2D Views of Multiple Objects](https://mlsanity.com/paper/1612.05872/) (2016)
- [Towards Unsupervised Learning of Generative Models for 3D Controllable Image Synthesis](https://mlsanity.com/paper/1912.05237/) (2019)
- [PanoHead: Geometry-Aware 3D Full-Head Synthesis in 360circ](https://mlsanity.com/paper/2303.13071/) (2023)
- [Lifting 2D StyleGAN for 3D-Aware Face Generation](https://mlsanity.com/paper/2011.13126/) (2020)
- [Unsupervised Generative 3D Shape Learning from Natural Images](https://mlsanity.com/paper/1910.00287/) (2019)
- [GGHead: Fast and Generalizable 3D Gaussian Heads](https://mlsanity.com/paper/2406.09377/) (2024)
- [Harnessing Diffusion-Yielded Score Priors for Image Restoration](https://mlsanity.com/paper/2507.20590/) (2025)
- [SphereHead: Stable 3D Full-head Synthesis with Spherical Tri-plane Representation](https://mlsanity.com/paper/2404.05680/) (2024)
- [LPFF: A Portrait Dataset for Face Generators Across Large Poses](https://mlsanity.com/paper/2303.14407/) (2023)
- [Portrait3D: Text-Guided High-Quality 3D Portrait Generation Using Pyramid Representation and GANs Prior](https://mlsanity.com/paper/2404.10394/) (2024)
- [15M Multimodal Facial Image-Text Dataset](https://mlsanity.com/paper/2407.08515/) (2024)
- [DiffPortrait360: Consistent Portrait Diffusion for 360 View Synthesis](https://mlsanity.com/paper/2503.15667/) (2025)
- [SOAP: Style-Omniscient Animatable Portraits](https://mlsanity.com/paper/2505.05022/) (2025)
- [DAViD: Data-efficient and Accurate Vision Models from Synthetic Data](https://mlsanity.com/paper/2507.15365/) (2025)
- [3DGH: 3D Head Generation with Composable Hair and Face](https://mlsanity.com/paper/2506.20875/) (2025)
- [HyPlaneHead: Rethinking Tri-plane-like Representations in Full-Head Image Synthesis](https://mlsanity.com/paper/2509.16748/) (2025)
- [VGGHeads: 3D Multi Head Alignment with a Large-Scale Synthetic Dataset](https://mlsanity.com/paper/2407.18245/) (2024)
- [AvatarBack: Back-Head Generation for Complete 3D Avatars from Front-View Images](https://mlsanity.com/paper/2508.20623/) (2025)

## Built on by (most cited first, 2 of 2)

- [Giving Faces Their Feelings Back: Explicit Emotion Control for Feedforward Single-Image 3D Head Avatars](https://mlsanity.com/paper/2604.14541/) (2026)
- [GeoFace: Consistent Multi-View Face Generation with Geometry-Constrained Diffusion](https://mlsanity.com/paper/2606.27659/) (2026)

## Abstract (arXiv)

Conditioning is crucial for stable training of full-head 3D GANs. Without any conditioning signal, the model suffers from severe mode collapse, making it impractical to training. However, a series of previous full-head 3D GANs conventionally choose the view angle as the conditioning input, which leads to a bias in the learned 3D full-head space along the conditional view direction. This is evident in the significant differences in generation quality and diversity between the conditional view and non-conditional views of the generated 3D heads, resulting in global incoherence across different head regions. In this work, we propose to use view-invariant semantic feature as the conditioning input, thereby decoupling the generative capability of 3D heads from the viewing direction. To construct a view-invariant semantic condition for each training image, we create a novel synthesized head image dataset. We leverage FLUX.1 Kontext to extend existing high-quality frontal face datasets to a wide range of view angles. The image clip feature extracted from the frontal view is then used as a shared semantic condition across all views in the extended images, ensuring semantic alignment while eliminating directional bias. This also allows supervision from different views of the same subject to be consolidated under a shared semantic condition, which accelerates training and enhances the global coherence of the generated 3D heads. Moreover, as GANs often experience slower improvements in diversity once the generator learns a few modes that successfully fool the discriminator, our semantic conditioning encourages the generator to follow the true semantic distribution, thereby promoting continuous learning and diverse generation. Extensive experiments on full-head synthesis and single-view GAN inversion demonstrate that our method achieves significantly higher fidelity, diversity, and generalizability.