[1]朱伟明,于家蓓.基于微调扩散模型的智能化衬衫款式图生成[J].服装学报,2025,10(03):260-267.
　ZHU Weiming,YU Jiabei.Intelligent Shirt Style Drawing Generation Based on Fine-Tuned Diffusion Model[J].Journal of Clothing Research,2025,10(03):260-267.
点击复制

基于微调扩散模型的智能化衬衫款式图生成()

《服装学报》[ISSN:2096-1928/CN:32-1864/TS]

卷:
第10卷
期数:
2025年03期
页码:
260-267
栏目:
图像处理及生成技术专题
出版日期:
2025-07-01

文章信息/Info

Title:
Intelligent Shirt Style Drawing Generation Based on Fine-Tuned Diffusion Model
作者:
朱伟明;  于家蓓
浙江理工大学 服装学院,杭州 310018
Author(s):
ZHU Weiming;  YU Jiabei
School of Fashion Design and Engineering, Zhejiang Sci-Tech University, Hangzhou 310018, China
分类号:
TP 391.41; TS 941.71
文献标志码:
A
摘要:
为提高服装设计产业的效率并降低成本,基于人工智能图像生成技术,通过低秩自适应微调Stable Diffusion大模型,实现端到端的款式图快速生成。以衬衫为例,通过分析扩散模型,构建衬衫款式图数据集并进行模型训练,同时采用结构相似性算法和峰值信噪比对生成效果进行量化评估。实验结果表明,微调后的模型在结构相似性算法和峰值信噪比指标上分别提升了22.69%和68.59%,且生成的款式图在结构、构图、风格、细节及线条质量上均显著优于微调前。该方法为服装智能设计提供了高效、低成本的新思路,具备实际推广潜力。
Abstract:
This study aims to enhance the efficiency and reduce costs in the apparel design industry by leveraging artificial intelligence image generation technology. An end-to-end rapid generation of garment style drawings is achieved through LoRA fine-tuning of the Stable Diffusion model. Taking shirts as the case study, the technical principles of diffusion models were analyzed. A dedicated shirt style drawing dataset was constructed and utilized for model training. The generation performance was quantitatively evaluated using the structural similarity index(SSIM)and peak signal-to-noise ratio(PSNR). Experimental results demonstrate that the fine-tuned model achieves 22.69% and 68.59% improvements in SSIM and PSNR metrics, respectively. The generated style drawings significantly outperform the baseline model in structural coherence, composition, style consis-tency, detail fidelity, and line quality.This method provides an efficient and cost-effective approach for intelligent garment design with substantial potential for practical implementation.

参考文献/References:

[1] 吴琼. 人工智能时代的创新设计思维[J]. 装饰, 2019(11): 18-21.
WU Qiong. Innovative design thinking in the era of artificial intelligence[J]. Zhuangshi, 2019(11): 18-21.(in Chinese)
[2] 詹希旎, 李白杨, 孙建军. 数智融合环境下AIGC的场景化应用与发展机遇[J]. 图书情报知识, 2023, 40(1): 75-85, 55.
ZHAN Xini, LI Baiyang, SUN Jianjun. Application scenarios and development opportunities of AIGC in the digital intelligence integration environment[J]. Documentation, Information and Know-ledge, 2023, 40(1): 75-85, 55.(in Chinese)
[3] 周楚轶, 柴春雷, 杨程. 人工智能辅助设计迭代研究——以平面设计为例[J]. 包装工程, 2021, 42(18): 50- 62.
ZHOU Chuyi, CHAI Chunlei, YANG Cheng. Artificial intelligence augmented design interation support: take graphic design for example[J]. Packaging Engineering, 2021, 42(18): 50- 62.(in Chinese)
[4] WU J J, ZHANG C K, XUE T F et al. Learning a probabilistic latent space of object shapes via 3D generative-adversarial modeling[J]. Advances in neural information processing systems,2016:29.
[5] ALMAHAIRI A, RAJESHWAR S, SORDONI A, et al. Augme-nted CycleGAN: Learning many-to-many mappings from unpaired data[C]//Proceedings of the 35th International Conference on Machine Learning(ICML). Stockholm:PMLR, 2018: 195-204.
[6] 李敏,刘冰清,彭庆龙等.基于CycleGAN算法的迷彩服装图案设计方法研究[J].丝绸,2022,59(8):100-106.
LI Min, LIU Bingqing, PENG Qinglong, et al. A camouflage suit pattern design based on the CycleGAN algorithm [J]. Journal of Silk,2022,59(8):100-106.(in Chinese)
[7] 贾静, 曹竟文, 徐平华, 等. 基于京剧脸谱意象色彩的服饰纹样自动配色[J]. 纺织学报, 2022, 43(12): 160-166.
JIA Jing, CAO Jingwen, XU Pinghua, et al. Automatic coloration of clothing pattern based on color parsing of Peking Opera masks[J]. Journal of Textile Research, 2022, 43(12): 160-166.(in Chinese)
[8] TANG H, GENG G H, ZHOU M Q. Application of digital processing in relic image restoration design[J]. Sensing and Imaging, 2020(21): 6.
[9] HO J, JAIN A, ABBEEL P. Denoising diffusion probabilistic models[J]. Advances in neural information processing systems,2020,33:6840- 6851.
[10] KARAGOZ H F, BAYKAL G, EKSI I A, et al. Textile Pattern Generation Using Diffusion Models[J]. ArXiv Preprint Server,2023:2304.
[11] ROMBACH R, BLATTMANN A, LORENZ D, et al. High-resolution image synthesis with latent diffusion models[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). New Orleans: IEEE, 2022: 10674-10685.
[12] ZHANG Q,CHEN M, BUKHARIN A, et al. Adaptive budget allocation for parameter-efficient fine-tuning[J]. ArXiv Preprint Server,2023:10512.
[13] HU E J, SHEN Y, WALLIS P, et al. Lora: Low-rank adaptation of large language models[J]. ArXiv Preprint Server, 2021:2106.
[14] HE J, ZHOU C, MA X, et al. Towards a unified view of parameter-efficient transfer learning[J]. ArXiv Preprint Server,2021:211.
[15] 肖红梅,王伟珍,房 媛.基于生成对抗网络的女上装图像属性编辑[J].服装学报,2024,9(1):42- 47.
XIAO Hongmei,WANG Weizhen,FANG Yuan.Image attribute editing of women’s tops based on generating adversarial networks[J].Journal of Clothing Research,2024,9(1):42- 47.(in Chinese)
[16] 陈雨琪, 薛涛, 刘俊华. 基于边缘增强和关联损失的服装图像风格迁移[J]. 现代纺织技术, 2024, 32(8): 117-126.
CHEN Yuqi, XUE Tao, LIU Junhua. Clothing pattern style transfer based on edge enhancement and association loss[J]. Advanced Textile Technology, 2024, 32(8): 117-126.(in Chinese)
(责任编辑:张 雪)

相似文献/References:

[1]杨道俊,沈 雷*,杨芳宇.基于扩散模型的生成式人工智能服装设计[J].服装学报,2025,10(04):311.
　YANG Daojun,SHEN Lei*,YANG Fangyu.Diffusion Model-Based Generative Artificial Intelligence Generated Content for Fashion Design[J].Journal of Clothing Research,2025,10(03):311.

更新日期/Last Update: 2025-06-30