TryOnGAN

Lewis, Kathleen M.; Varadharajan, Srivatsan; Kemelmacher-Shlizerman, Ira

doi:10.1145/3450626.3459884

Cited by 59 publications

(15 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Figure . 6 shows the result of image editing by distilling five interpretable directions from the eigenvectors of 𝑨 𝑇𝑜𝑅𝐺𝐵 ⊺ 𝑨 𝑻𝒐𝑹𝑮𝑩 where 𝛼 𝑖 (𝑖 ∈ [1,5]) means the intensity of each optimal directions 𝒏 𝒊 * in Equation. (12). These results demonstrate that we can edit the image color by manually adapting 𝛼 𝑖 while maintaining the structure and the texture of the generated clothing images thanks to the second module of our system.…”

Section: Quality Of Sefa Color Editingmentioning

confidence: 56%

“…One of the most promising methods is StyleGAN [9], which improves PGGAN [10] by using Adaptive Instance Normalization (AdaIN) [11] to generate realistic high-resolution images. StyleGAN image generation tasks can be applied to fashion image generation techniques such as style transfer-based virtual try-on [12] and fashion outfit generation [13]. StyleGAN2 [14] improves the original StyleGAN's training stability by refining AdaIN and introducing a lazy path length regularization method.…”

Section: Application Of Generative Adversarial Network In Fashionmentioning

confidence: 99%

See 1 more Smart Citation

GAN-based Detailed Clothing Generation System

Masukawa

Haji²,

Takagi³

et al. 2023

IIAI Letters on Informatics and Interdisciplinary Research

View full text Add to dashboard Cite

The implementation of Generative Adversarial Networks (GAN) in the fashion domain has been researched for various applications such as virtual-try-on, fashion item recommendation, and design generation. In this paper, we propose a GAN-based fashion design generation system that reduces the workload of the labor-intensive design creation task. Our system consists of two generative models: one that produces images of fashion items without any clothing patterns using conditioned StyleGAN2-ADA, and one that is a style transfer model reflecting the fine texture of the fashion item. The system also allows users to edit images of garments by manipulating the latent code of the generator. We demonstrate through qualitative and quantitative experiments that the proposed system trained on a dataset of real clothing inventory images can generate realistic and diverse images that reflect the input conditions in detail.

show abstract

Section: Quality Of Sefa Color Editingmentioning

confidence: 56%

Section: Application Of Generative Adversarial Network In Fashionmentioning

confidence: 99%

GAN-based Detailed Clothing Generation System

Masukawa

Haji²,

Takagi³

et al. 2023

IIAI Letters on Informatics and Interdisciplinary Research

View full text Add to dashboard Cite

show abstract

“…They then use the two networks to train a segmentation mask generation network in an unsupervised manner. This notion, of extracting a segmentation map with the help of StyleGAN's structure, has been employed similarly by others [ZLG*21,LY K *21,LKL*21,LVKS21].…”

Section: Discriminative Applicationsmentioning

confidence: 99%

“…While all the aforementioned works showed incredible results and promise in real‐world scenarios, they are limited in the domains they operate over. Some works have explored going beyond the facial domain and have explored applying StyleGAN for full‐body synthesis in various applications such as virtual try‐on and portrait reposing [LVKS21,ALY*21].…”

Section: Encoding and Inversionmentioning

confidence: 99%

State‐of‐the‐Art in the Architecture, Methods and Applications of StyleGAN

Bermano

Gal

Alaluf

et al. 2022

Computer Graphics Forum

View full text Add to dashboard Cite

Generative Adversarial Networks (GANs) have established themselves as a prevalent approach to image synthesis. Of these, StyleGAN offers a fascinating case study, owing to its remarkable visual quality and an ability to support a large array of downstream tasks. This state-of-the-art report covers the StyleGAN architecture, and the ways it has been employed since its conception, while also analyzing its severe limitations. It aims to be of use for both newcomers, who wish to get a grasp of the field, and for more experienced readers that might benefit from seeing current research trends and existing tools laid out. Among StyleGAN's most interesting aspects is its learned latent space. Despite being learned with no supervision, it is surprisingly well-behaved and remarkably disentangled. Combined with StyleGAN's visual quality, these properties gave rise to unparalleled editing capabilities. However, the control offered by StyleGAN is inherently limited to the generator's learned distribution, and can only be applied to images generated by StyleGAN itself. Seeking to bring StyleGAN's latent control to real-world scenarios, the study of GAN inversion and latent space embedding has quickly gained in popularity. Meanwhile, this same study has helped shed light on the inner workings and limitations of StyleGAN. We map out StyleGAN's impressive story through these investigations, and discuss the details that have made StyleGAN the go-to generator. We further elaborate on the visual priors StyleGAN constructs, and discuss their use in downstream discriminative tasks. Looking forward, we point out StyleGAN's limitations and speculate on current trends and promising directions for future research, such as task and target specific fine-tuning.

show abstract

“…Technically, there are two general approaches for controlling the images’ generation features under continuous conditions. One approach achieves attribute control by identifying operable paths in the latent space, , for instance, the separation hyperplane of a binary attribute, since the normal line of the hyperplane indicates a unique direction that it can be used to control the classification attribute. However, it can only provide relative control (e.g., turning the face older or rotating the face toward the left) without providing explicit control.…”

Section: Introductionmentioning

confidence: 99%

Development of an Improved One-Hot Encoding Method for Bubbly Flow Image Prediction Generation under Continuous Superficial Gas Velocities

Zhong

Ruan

Qian

et al. 2023

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

In recent years, bubble segmentation and detection algorithms have significantly contributed to the study of two-phase interface parameters. Nevertheless, high-quality bubbly flow image data acquisition remains a major constraint to their development. The application of deep learning methods, especially Conditional StyleGAN2, has alleviated this problem to some extent. However, it is restricted to categorical conditions, which means it could only generate bubbly flow images at superficial gas velocities included in the training data. In this work, we propose an improved one-hot encoding method to extend our previous work to predict and synthesize bubbly flow images under continuous superficial gas velocity. A reference sample set (RSS) is constructed and encoded at first to train the StyleGAN2 model, and the linear interpolation method is then employed by referring to the RSS to encode sample sets with arbitrary flow velocities and to generate the corresponding bubbly flow images. The proposed method is validated by training on an RSS which includes 9 real superficial gas velocities to generate images at different superficial gas velocities (include or exclude in the training data). Interface parameters, such as bubble size distribution, average equivalent diameter, void fraction, and interfacial area concentration, are adopted as quantitative evaluation metrics to evaluate the model's synthesis and prediction ability. The comparative analysis results indicate that the proposed method can generate bubbly flow images at an arbitrary superficial gas velocity in the studied range with an absolute relative error of no more than 9% under all metrics, which is potentially helpful for the advancement of two-phase flow investigation in practical implementations.

show abstract

TryOnGAN

Cited by 59 publications

References 28 publications

GAN-based Detailed Clothing Generation System

GAN-based Detailed Clothing Generation System

State‐of‐the‐Art in the Architecture, Methods and Applications of StyleGAN

Development of an Improved One-Hot Encoding Method for Bubbly Flow Image Prediction Generation under Continuous Superficial Gas Velocities

Contact Info

Product

Resources

About