计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (5): 207-217.doi: 10.11896/jsjkx.251100057

• 计算机图形学 & 多媒体 • 上一篇    下一篇

uHairDet:面向雄激素性脱发诊断的仿真毛发图像生成与检测方法

陈琦1,2, 陈星凯1,2, 张辉煌1,2, 苏一平3, 胡海根1,2   

  1. 1 浙江工业大学计算机科学与技术学院 杭州 310023
    2 全省可视信息智能处理重点实验室 杭州 310023
    3 杭州市临安区第一人民医院皮肤科 杭州 311300
  • 收稿日期:2025-11-12 修回日期:2026-02-11 发布日期:2026-05-08
  • 通讯作者: 胡海根(hghu@zjut.edu.cn)
  • 作者简介:(chenqi@zjut.edu.cn)
  • 基金资助:
    国家自然科学基金(62373324); 浙江省尖兵领雁项目(2025C02160,2026C02A1221)

uHairDet:Method for Synthetic Hair Image Generation and Detection in Androgenetic AlopeciaDiagnosis

CHEN Qi1,2, CHEN Xingkai1,2, ZHANG Huihuang1,2, SU Yiping3, HU Haigen1,2   

  1. 1 School of Computer Science and Technology, Zhejiang University of Technology, Hangzhou 310023, China
    2 Zhejiang Key Laboratory of Visual Information Intelligent Processing, Zhejiang University of Technology, Hangzhou 310023, China
    3 Department of Dermatology, The First People’s Hospital of Lin’an District, Hangzhou 311300, China
  • Received:2025-11-12 Revised:2026-02-11 Online:2026-05-08
  • About author:CHEN Qi,born in 1977,Ph.D.His main research interests include machine learning and intelligent information systems.
    HU Haigen,born in 1977,Ph.D,professor.His main research interests include computer vision,machine learning and medical image processing.
  • Supported by:
    National Natural Science Foundation of China(62373324) and Zhejiang Province Leading Geese Plan(2025C02160,2026C02A1221).

PDF (PC)

253

摘要： 雄激素性脱发的临床诊断主要依赖于人工毛发计数,该方法不仅繁琐耗时,且主观性强,其严重制约了临床诊断的效率与一致性。因此,开发能够实现毛发自动识别与计数的智能辅助诊断技术具有重要的临床意义。然而,由于毛发结构纤细,且缺乏像素级标注的高质量训练数据,现有的监督检测方法仍面临巨大挑战。为应对该挑战,提出了一种名为 uHairDet 的毛发检测新方法,其核心思路是通过合成具有标注信息的毛发图像来降低对手工标注数据的依赖。该方法包含3个部分:1)提出了自标注毛发数据生成器(HBDG),用于合成具有像素级标注信息、结构合理的毛发图像;2)提出了结合语义感知自适应纠错损失(HEE Loss) 的结构稳定风格迁移模型(BS-GAN),以提升在风格迁移过程中图像结构的稳定性;3)引入了基于Mean-Teacher的框架的FCOS+PSC检测模型,配合提出的适用于毛发的OBB标注范式对模型进行训练,解决了传统毛发标注范式丢失大量毛发有效信息的问题。结果表明,在完全无须手工标注数据的情况下,该方法取得了 56.9% 的平均精度,显著优于多种基线模型,有望为皮肤科毛发疾病的智能辅助诊断建立一种新范式。

关键词: 毛发检测与计数, 旋转目标检测, 域适应, 毛发生成, 损失函数

Abstract: The clinical diagnosis of Androgenetic Alopecia(AGA) relies heavily on manual hair counting,a process that is time-consuming,subjective,and limits diagnostic efficiency.While automated hair recognition technology holds great promise,it faces challenges due to fine hair structures and a lack of pixel-level annotated data.To address this,this paper proposes uHairDet,a novel framework that minimizes dependency on manual annotations by leveraging synthetically generated data.The proposed approach consists of three key components:1)a hair synthesis with built-in annotations data generator(HBDG) for creating structurally plausible hair images with pixel-level labels;2)a structure-stable style-transfer GAN(BS-GAN) incorporating a semantic-aware adaptive error correction loss(HEE Loss) to enhance structural consistency;3)an FCOS+PSC detection model based on a Mean-Teacher framework,trained with a new oriented bounding box(OBB) annotation paradigm tailored for hairs to preserve critical information.Results demonstrate that the proposed method,requiring no manual labels,achieves a 56.9% AP,significantly outperforming baseline models and establishing a new paradigm for intelligent assisted diagnosis in hair-related disorders.

Key words: Hair detection and counting, Rotation detection, Domain adaptation, Hair synthesis, Loss function

中图分类号: 

  • TP391
[1]KIM M,KANG S,LEE B D.Evaluation of automated measurement of hair density using deep neural networks[J].Sensors,2022,22(2):650.
[2]KIM J H,KWON S,FU J,et al.Hair follicle classification and hair loss severity estimation using mask R-CNN[J].Journal of Imaging,2022,8(10):283.
[3]WANG Y F,HSU M H,WANG M Y F.Comparative analysis of automatic hair count estimation methods[J].Archives of Dermatological Research,2025,317(1):361.
[4]CHEN M C,CHANG W J,XIAO Y X,et al.A hair volumeanalysis system using deep learning and image processing technologies[C]//2021 IEEE International Conference on Consumer Electronics-Taiwan(ICCE－TW).IEEE,2021:1-2.
[5]BOCHKOVSKIY A,WANG C Y,LIAO H Y M.Yolov4:Opti-mal speed and accuracy of object detection[J].arXiv:2004.10934,2020.
[6]ZHU J Y,PARK T,ISOLA P,et al.Unpaired image-to-imagetranslation using cycle-consistent adversarial networks[C]//Proceedings of the IEEE International Conference on Computer Vision.2017:2223-2232.
[7]WANG X,YU J.Learning to cartoonize using white-box cartoon representations[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:8090-8099.
[8]CHOI Y,UH Y,YOO J,et al.v2:Diverse image synthesis for multiple domains.In 2020 IEEE[C]//CVF Conference on Computer Vision and Pattern Recognition(CVPR).2020:8185-8194.
[9]PARK T,EFROS A A,ZHANG R,et al.Contrastive learning for unpaired image-to-image translation[C]//European Confe-rence on Computer Vision.Cham:Springer,2020:319-345.
[10]YI R,LIU Y J,LAI Y K,et al.Unpaired portrait drawing generation via asymmetric cycle mapping[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:8217-8225.
[11]TARVAINEN A,VALPOLA H.Mean teachers are better role models:Weight-averaged consistency targets improve semi-supervised deep learning results[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems.2017:1195-1204.
[12]SHIH H C.An unsupervised hair segmentation and countingsystem in microscopy images[J].IEEE Sensors Journal,2014,15(6):3565-3572.
[13]JÄHNE B.Digital image processing[M].Berlin:Springer,2005.
[14]SHIH H C,LIU E R.Adaptive region merging approach formorphological color image segmentation[C]//Proceedings of the ACCV.2014:1-14.
[15]OTSU N.A threshold selection method from gray-level histograms[J].Automatica,1975,11(285-296):23-27.
[16]ZHANG Q,EUN S J.Design and implementation of an automatic hair counting system[J].Journal of Digital Art Enginee-ring and Multimedia,2014,1(2):75.
[17]HOUGH P V C,POWELL B W.A method for faster analysis of bubble chamber photographs[J].Il Nuovo Cimento(1955-1965),1960,18(6):1184-1191.
[18]KIM W,KIM H,REW J,et al.A hair density measuring scheme using smartphone[C]//Annual Conference of KIPS.Korea Information Processing Society,2015:1416-1419.
[19]GATYS L A,ECKER A S,BETHGE M.A neural algorithm of artistic style[J].arXiv:1508.06576,2015.
[20]HUANG X,BELONGIE S.Arbitrary style transfer inreal-time with adaptive instance normalization[C]//Proceedings of the IEEE International Conference on Computer Vision.2017:1501-1510.
[21]GOODFELLOW I J,POUGET-ABADIE J,MIRZA M,et al.Generative adversarial nets[C]//Proceedings of the 28st International Conference on Neural Information Processing Systems.2014.
[22]CHO H,LEE J,CHANG S,et al.One-shot structure-awarestylized image synthesis[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2024:8302-8311.
[23]WANG Z,ZHAO L,XING W.Stylediffusion:Controllable disentangled style transfer via diffusion models[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.2023:7677-7689.
[24]EVERAERT M N,BOCCHIO M,ARPA S,et al.Diffusion instyle[C]//Proceedings of the IEEE/CVF International Confe-rence on Computer Vision.2023:2251-2261.
[25]YANG S,HWANG H,YE J C.Zero-shot contrastive loss fortext-guided diffusion image style transfer[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.2023:22873-22882.
[26]KWON G,YE J C.Diffusion-based image translation using disentangled style and content representation[J].arXiv:2209.15264,2022.
[27]ZHANG Y,HUANG N,TANG F,et al.Inversion-based creativity transfer with diffusion models[J].arXiv:2211.13203,2022.
[28]ZHANG Y,DONG W,TANG F,et al.Prospect:Expanded conditioning for the personalization of attribute-aware image gene-ration[C]//CoRR.2023.
[29]MOU C,WANG X,XIE L,et al.T2i-adapter:Learning adapters to dig out more controllable ability for text-to-image diffusion models[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2024:4296-4304.
[30]ZHANG Z,ZHANG Q,XING W,et al.Artbank:Artistic style transfer with pre-trained diffusion model and implicit style prompt bank[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2024:7396-7404.
[31]DALAL N,TRIGGS B.Histograms of oriented gradients forhuman detection[C]//2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition(CVPR’05).IEEE,2005:886-893.
[32]LOWE D G.Distinctive image features from scale-invariant keypoints[J].International Journal of Computer Vision,2004,60(2):91-110.
[33]MA J,SHAO W,YE H,et al.Arbitrary-oriented scene text detection via rotation proposals[J].IEEE Transactions on Multimedia,2018,20(11):3111-3122.
[34]YANG X,ZHOU Y,ZHANG G,et al.The KFIoU loss for rotated object detection[J].arXiv:2201.12558,2022.
[35]TIAN Z,SHEN C,CHEN H,et al.FCOS:A simple and strong anchor-free object detector[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2020,44(4):1922-1933.
[36]YU Y,DA F.Phase-shifting coder:Predicting accurate orientation in oriented object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2023:13354-13363.
[37]GANIN Y,USTINOVA E,AJAKAN H,et al.Domain-adversarial training of neural networks[J].Journal of Machine Lear-ning Research,2016,17(59):1-35.
[38]GANIN Y,LEMPITSKY V.Unsupervised domain adaptation by backpropagation[C]//International Conference on Machine Learning.PMLR,2015:1180-1189.
[39]ZHANG W,OUYANG W,LI W,et al.Collaborative and adversarial network for unsupervised domain adaptation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2018:3801-3809.
[40]PEI Z,CAO Z,LONG M,et al.Multi-adversarial domain adaptation[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2018.
[41]TZENG E,HOFFMAN J,ZHANG N,et al.Deep domain confusion:Maximizing for domain invariance[J].arXiv:1412.3474,2014.
[42]LONG M,CAO Y,WANG J,et al.Learning transferable fea-tures with deep adaptation networks[C]//International Confe-rence on Machine Learning.PMLR,2015:97-105.
[43]LONG M,ZHU H,WANG J,et al.Unsupervised domain adaptation with residual transfer networks[C]//Proceedings of the 30st International Conference on Neural Information Processing Systems.2016:136-144.
[44]LONG M,ZHU H,WANG J,et al.Deep transfer learning with joint adaptation networks[C]//International Conference on Machine Learning.PMLR,2017:2208-2217.
[45]GHIFARY M,KLEIJN W B,ZHANG M,et al.Deep recon-struction-classification networks for unsupervised domain adaptation[C]//European Conference on Computer Vision.Cham:Springer,2016:597-613.
[46]BOUSMALIS K,TRIGEORGIS G,SILBERMAN N,et al.Domain separation networks[C]//NIPS 2016.2016.
[47]LI Y J,DAI X,MA C Y,et al.Cross-domain adaptive teacher for object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2022:7581-7590.
[48]KENNERLEY M,WANG J G,VEERAVALLI B,et al.2pcnet:Two-phase consistency training for day-to-night unsupervised domain adaptive object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2023:11484-11493.
[49]LIN T Y,GOYAL P,GIRSHICK R,et al.Focal loss for dense object detection[C]//Proceedings of the IEEE International Conference on Computer Vision.2017:2980-2988.
[50]YANG Z,LIU S,HU H,et al.Reppoints:Point set representation for object detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.2019:9657-9666.
[51]LI W,CHEN Y,HU K,et al.Oriented reppoints for aerial object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2022:1829-1838.
[52]HOU L,LU K,XUE J,et al.Shape-adaptive selection and mea-surement for oriented object detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2022:923-932.
[53]WEN L,GAO C,ZOU C.CAP-VSTNet:Content affinity preserved versatile style transfer[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2023:18300-18309.
[54]ZHOU Y,YANG X,ZHANG G,et al.Mmrotate:A rotated object detection benchmark using pytorch[C]//Proceedings of the 30th ACM International Conference on Multimedia.2022:7331-7334.
[55]XIA G S,BAI X,DING J,et al.DOTA:A large-scale dataset for object detection in aerial images[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2018:3974-3983.
[56]ZHANG R,ISOLA P,EFROS A A,et al.The unreasonable effectiveness of deep features as a perceptual metric[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2018:586-595.
[57]HEUSEL M,RAMSAUER H,UNTERTHINER T,et al.GANs trained by a two time-scale update rule converge to a local nash equilibrium[C]//NIPS 2017.2017.
[58]BROOKS T,HOLYNSKI A,EFROS A A.Instructpix2pix:Learning to follow image editing instructions[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2023:18392-18402.
[59]XU Z.SkinDualGen:Prompt-Driven Diffusion for Simultaneous Image-Mask Generation in Skin Lesions[J].arXiv:2507.19970,2025.
[60]CAO S,JOSHI D,GUI L Y,et al.Contrastive mean teacher for domain adaptive object detectors[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2023:23839-23848.
[61]TRANHEDEN W,OLSSON V,PINTO J,et al.Dacs:Domainadaptation via cross-domain mixed sampling[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision.2021:1379-1389.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发