The Devil Is in the Leakage: A Disentangled Dual-Purification Framework for High-Fidelity Hairstyle Transfer

• URL: http://arxiv.org/abs/2607.11281v1
• Date: Mon, 13 Jul 2026 09:00:42 GMT
• Title: The Devil Is in the Leakage: A Disentangled Dual-Purification Framework for High-Fidelity Hairstyle Transfer
• Abstract summary: Hairstyle transfer aims to synthesize a portrait by transplanting the hairstyle from a reference image onto a source subject while preserving the source identity.<n>Recent foundation models show strong generative capability, but they struggle with the zero-shot disentanglement required for precise local editing.<n>We propose the Dual-Purification Framework (DPF), which introduces two complementary training-time regularizers.
• Score: 19.467144465203802
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Hairstyle transfer aims to synthesize a photorealistic portrait by transplanting the hairstyle from a reference image onto a source subject while preserving the source identity. Recent foundation models show strong generative capability, but they struggle with the zero-shot disentanglement required for precise local editing, often entangling the reference hairstyle with its original identity and pose. Existing diffusion-based pipelines typically decompose the task by first generating a "bald" image from the source and then injecting hairstyle features from the reference. However, we show that this paradigm suffers from a fundamental leakage problem. Identity Leakage in Hairstyle occurs when hairstyle features retain reference identity or pose information, while Flaw Leakage in Bald arises when residual artifacts in the bald image are propagated into the final synthesis. To address both issues, we propose the Dual-Purification Framework (DPF), which introduces two complementary training-time regularizers. Adversarial Hairstyle Purification (AHP) purifies hairstyle features by suppressing identity predictability under a mutual-information-inspired adversarial objective. Contrastive Geometric Purification (CGP) regularizes the ControlNet pathway with a contrastive objective, reducing the model's reliance on geometric artifacts in the bald condition. By jointly purifying the hairstyle representation and geometric pathway, DPF achieves high-fidelity, identity-preserving hairstyle transfer and state-of-the-art performance on diverse benchmarks.

Related papers

• H-Adapter: Pose-Robust Hairstyle Transfer via Attention-Derived, Source-Aligned Hair Masks [2.9567293946666173]
  H-Adapter improves pose robustness by training with a region-specific loss that disentangles hair and non-hair objectives.<n> Experiments on pose-agnostic and pose-different subsets demonstrate strong quantitative results.<n>We introduce a VLM-as-a-judge protocol and observe consistent gains in hairstyle faithfulness, non-hair preservation, and artifact quality.
  arXiv  Detail & Related papers  (2026-06-24T08:52:06Z)
• HairPort: In-context 3D-aware Hair Import and Transfer for Images [12.811554431300996]
  HairPort is a 3D-aware hairstyle transfer framework.<n>It produces realistic bald versions of faces through LoRA-based in-context adaptation of FLUX.1 Kontext.<n>Being 3D aware, our method supports large pose and scale discrepancies between the source and target.
  arXiv  Detail & Related papers  (2026-06-10T18:13:18Z)
• HairGPT: Strand-as-Language Autoregressive Modeling for Realistic 3D Hairstyle Synthesis [62.89988554372163]
  HairGPT is a strand-centric framework that treats strands as generative primitives.<n>It formulates realistic 3D hairstyle synthesis as a dual-decoupled autoregressive sequence modeling problem.
  arXiv  Detail & Related papers  (2026-05-09T09:19:36Z)
• HairGS: Hair Strand Reconstruction based on 3D Gaussian Splatting [50.93221272778306]
  Human hair reconstruction is a challenging problem in computer vision.<n>We extend the 3DGS framework to enable strand-level hair geometry reconstruction from multi-view images.<n>Our method robustly handles a wide range of hairstyles and achieves efficient reconstruction, typically completing within one hour.
  arXiv  Detail & Related papers  (2025-09-09T14:08:41Z)
• Im2Haircut: Single-view Strand-based Hair Reconstruction for Human Avatars [60.99229760565975]
  We present a novel approach for 3D hair reconstruction from single photographs based on a global hair prior combined with local optimization.<n>We exploit this prior to create a Gaussian-splatting-based reconstruction method that creates hairstyles from one or more images.
  arXiv  Detail & Related papers  (2025-09-01T13:38:08Z)
• DiffLocks: Generating 3D Hair from a Single Image using Diffusion Models [53.08138861924767]
  We propose DiffLocks, a novel framework that enables reconstruction of a wide variety of hairstyles directly from a single image.<n>First, we address the lack of 3D hair data by automating the creation of the largest synthetic hair dataset to date, containing 40K hairstyles.<n>By using a pretrained image backbone, our method generalizes to in-the-wild images despite being trained only on synthetic data.
  arXiv  Detail & Related papers  (2025-05-09T16:16:42Z)
• Realistic and Efficient Face Swapping: A Unified Approach with Diffusion Models [69.50286698375386]
  We propose a novel approach that better harnesses diffusion models for face-swapping. We introduce a mask shuffling technique during inpainting training, which allows us to create a so-called universal model for swapping. Ours is a relatively unified approach and so it is resilient to errors in other off-the-shelf models.
  arXiv  Detail & Related papers  (2024-09-11T13:43:53Z)
• What to Preserve and What to Transfer: Faithful, Identity-Preserving Diffusion-based Hairstyle Transfer [35.80645300182437]
  Existing hairstyle transfer approaches rely on StyleGAN.<n>We propose a one-stage hairstyle transfer diffusion model, HairFusion, that applies to real-world scenarios.<n>Our method achieves state-of-the-art performance compared to the existing methods in preserving the integrity of both the transferred hairstyle and the surrounding features.
  arXiv  Detail & Related papers  (2024-08-29T11:30:21Z)
• HairStep: Transfer Synthetic to Real Using Strand and Depth Maps for Single-View 3D Hair Modeling [55.57803336895614]
  We tackle the challenging problem of learning-based single-view 3D hair modeling. We first propose a novel intermediate representation, termed as HairStep, which consists of a strand map and a depth map. It is found that HairStep not only provides sufficient information for accurate 3D hair modeling, but also is feasible to be inferred from real images.
  arXiv  Detail & Related papers  (2023-03-05T15:28:13Z)
• Style Your Hair: Latent Optimization for Pose-Invariant Hairstyle Transfer via Local-Style-Aware Hair Alignment [29.782276472922398]
  We propose a pose-invariant hairstyle transfer model equipped with latent optimization and a newly presented local-style-matching loss. Our model has strengths in transferring a hairstyle under larger pose differences and preserving local hairstyle textures.
  arXiv  Detail & Related papers  (2022-08-16T14:23:54Z)
• HairFIT: Pose-Invariant Hairstyle Transfer via Flow-based Hair Alignment and Semantic-Region-Aware Inpainting [26.688276902813495]
  We propose a novel framework for pose-invariant hairstyle transfer, HairFIT. Our model consists of two stages: 1) flow-based hair alignment and 2) hair synthesis. Our SIM estimator divides the occluded regions in the source image into different semantic regions to reflect their distinct features during the inpainting.
  arXiv  Detail & Related papers  (2022-06-17T06:55:20Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.