Abstract
Texture and shape in fashion, constituting essential elements of garments, characterize the body and surface of the fabric and outline the silhouette of clothing, respectively. The selection of texture and shape plays a critical role in the design process, as they largely determine the success of a new design for fashion items. In this research, we propose a texture and shape disentangled generative adversarial network (TSD-GAN) to perform “intelligent” design with the transformation of texture and shape in fashion items. Our TSD-GAN aims to learn how to disentangle the features of texture and shape of different fashion items in an unsupervised manner. Specifically, a fashion attribute encoder is developed to decompose the input fashion items into independent representations of texture and shape. Then, to learn the coarse or fine styles hidden in the features of texture and shape, a texture mapping network and a shape mapping network are proposed to disentangle the features into different hierarchical representations. The different hierarchical representations of texture and shape are then fed into a multi-factor-based generator to generate mixed-style fashion items. In addition, a multi-discriminator framework is developed to distinguish the authenticity and texture similarity between the generated images and the real images. Experimental results on different fashion categories demonstrate that our proposed TSD-GAN may be useful for assisting designers to accomplish the design process by transforming the texture and shape of fashion items.
- [1] . 2019. Image2stylegan: How to embed images into the StyleGan latent space. In Proceedings of the CVPR. 4431–4440.Google Scholar
Cross Ref
- [2] . 2020. Disentangled image generation through structured noise injection. In Proceedings of the CVPR. 5133–5141.Google Scholar
Cross Ref
- [3] . 2018. Virtual reality interactive teaching for Chinese traditional Tibetan clothing. Art, Design Commun. Higher Edu. 17, 1 (2018), 51–59.Google Scholar
Cross Ref
- [4] . 2019. Image-to-image translation via group-wise deep whitening-and-coloring transformation. In Proceedings of the CVPR. 10639–10647.Google Scholar
Cross Ref
- [5] . 2020. Stargan v2: Diverse image synthesis for multiple domains. In Proceedings of the CVPR. 8185–8194.Google Scholar
Cross Ref
- [6] . 2020. Editing in style: Uncovering the local semantics of GANs. In Proceedings of the CVPR. 5770–5779.Google Scholar
Cross Ref
- [7] . 2018. FashionGAN: Display your fashion design using conditional generative adversarial nets. In Computer Graphics Forum, Vol. 37. Wiley Online Library, 109–119.Google Scholar
- [8] . 2020. Arbitrary style transfer via multi-adaptation network. In Proceedings of the ACM MM. 2719–2727.Google Scholar
Digital Library
- [9] . 2020. Fashion editing with adversarial parsing learning. In Proceedings of the CVPR. 8117–8125.Google Scholar
Cross Ref
- [10] . 2020. Collaborative learning for faster stylegan embedding. Retrieved from https://arXiv:2007.01758.Google Scholar
- [11] . 2022. Personalized fashion compatibility modeling via metapath-guided heterogeneous graph learning. In Proceedings of the ACM SIGIR. 482–491.Google Scholar
Digital Library
- [12] . 2021. Multimodal compatibility modeling via exploring the consistent and complementary correlations. In Proceedings of the ACM MM. 2299–2307.Google Scholar
Digital Library
- [13] . 2020. Ganspace: Discovering interpretable gan controls. Adv. Neural Info. Process. Syst. 33 (2020), 9841–9850.Google Scholar
- [14] . 2016. Deep residual learning for image recognition. In Proceedings of the CVPR. 770–778.Google Scholar
Cross Ref
- [15] . 2017. Gans trained by a two time-scale update rule converge to a local nash equilibrium. Adv. Neural Info. Process. Syst. 30 (2017), 6626–6637.Google Scholar
- [16] . 2020. Sketch-guided deep portrait generation. ACM Trans. Multim. Comput. Commun. Appl. 16, 3 (2020), 88:1–88:18.Google Scholar
Digital Library
- [17] . 2018. Multimodal unsupervised image-to-image translation. In Proceedings of the ECCV. 179–196.Google Scholar
Digital Library
- [18] . 2018. Salient object detection via multi-scale attention CNN. Neurocomputing 322 (2018), 130–140.Google Scholar
Cross Ref
- [19] . 2021. Deep learning for fashion style generation. IEEE TNNLS (2021).Google Scholar
- [20] . 2016. Perceptual losses for real-time style transfer and super-resolution. In Proceedings of the ECCV. Springer, 694–711.Google Scholar
Cross Ref
- [21] . 2020. Analyzing and improving the image quality of stylegan. In Proceedings of the CVPR. 8107–8116.Google Scholar
Cross Ref
- [22] . 2020. Style-controlled synthesis of clothing segments for fashion image manipulation. IEEE TMM 22, 2 (2020), 298–310.Google Scholar
Digital Library
- [23] . 2019. Content and style disentanglement for artistic style transfer. In Proceedings of the ICCV. 4421–4430.Google Scholar
Cross Ref
- [24] . 2019. Technology for creating images in AutoCAD. Eur. J. Res. Reflect. Edu. Sci. 7 (2019).Google Scholar
- [25] . 2018. Diverse image-to-image translation via disentangled representations. In Proceedings of the ECCV. 36–52.Google Scholar
Digital Library
- [26] . 2013. Content-based image retrieval using color difference histogram. Pattern Recogn. 46, 1 (2013), 188–198.Google Scholar
Digital Library
- [27] . 2005. The CIEDE2000 color-difference formula: Implementation notes, supplementary test data, and mathematical observations. Color Res. Appl.: Endors. Inter-Soc. Color Council 30, 1 (2005), 21–30.Google Scholar
Cross Ref
- [28] . 2009. Adobe Photoshop CS4: Comprehensive Concepts and Techniques. Cengage Learning.Google Scholar
- [29] . 2012. Adobe Illustrator CS6 Classroom in a Book. Adobe Press.Google Scholar
Digital Library
- [30] . 2018. Texturegan: Controlling deep image synthesis with texture patches. In Proceedings of the CVPR. 8456–8465.Google Scholar
Cross Ref
- [31] . 2022. Toward intelligent design: An AI-based fashion designer using generative adversarial networks aided by sketch and rendering generators. IEEE TMM (2022).Google Scholar
- [32] . 2019. Photorealistic style transfer via wavelet transforms. In Proceedings of the ICCV. 9035–9044.Google Scholar
Cross Ref
- [33] . 2022. Tell, imagine, and search: End-to-end learning for composing text and image to image retrieval. ACM Trans. Multim. Comput. Commun. Appl. 18, 2 (2022), 59:1–59:23.Google Scholar
Digital Library
- [34] . 2018. The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the CVPR. 586–595.Google Scholar
Cross Ref
- [35] . 2022. Domain enhanced arbitrary image style transfer via contrastive learning. In Proceedings of the ACM SIGGRAPH.Google Scholar
Digital Library
- [36] . 2021. M6-ufc: Unifying multi-modal controls for conditional image synthesis. Retrieved from https://arXiv:2105.14211.Google Scholar
- [37] . 2020. Sean: Image synthesis with semantic region-adaptive normalization. In Proceedings of the CVPR. 5103–5112.Google Scholar
Cross Ref
- [38] . 2017. Be your own Prada: Fashion synthesis with structural coherence. In Proceedings of the ICCV. 1689–1697.Google Scholar
Cross Ref
Index Terms
Toward Intelligent Fashion Design: A Texture and Shape Disentangled Generative Adversarial Network
Recommendations
Shape from Texture Using Local Spectral Moments
We present a non-feature-based solution to the problem of computing the shape of curved surfaces from texture information. First, the use of local spatial-frequency spectra and their moments to describe texture is discussed and motivated. A new, more ...
Maximally Stable Texture Regions
ICPR '10: Proceedings of the 2010 20th International Conference on Pattern RecognitionIn this study, we propose to detect interest regions based on texture information of images. For this purpose, Maximally Stable Extremal Regions (MSER) approach is extended using the high dimensional texture features of image pixels. The regions with ...
Image retrieval by using texture and shape correlated hand crafted features
Content-based image retrieval (CBIR) has become one of the trending areas of research in computer vision. In this paper, consonance on hue, saturation, and intensity is used by applying inter-channel voting between them. Diagonally symmetric pattern (DSP) ...






Comments