Colorization of Logo Sketch Based on Conditional Generative Adversarial Networks

Tian, Nannan; Liu, Yuan; Wu, Bo; Li, Xiaofeng

doi:10.3390/electronics10040497

Cited by 10 publications

(9 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Early methods (Tian 2016(Tian , 2017Jiang et al 2017;Lyu et al 2017;Chang et al 2018;Sun, Zhang, and Yang 2018;Jiang et al 2019;Yang et al 2019a,b;Gao and Wu 2020;Wu, Yang, and Hsu 2020;Wen et al 2021;Hassan, Ahmed, and Choi 2021) utilize Image-to-Image translation networks (Zhang et al 2022) to achieve font generation by learning the mapping function between different fonts. Tian et al presents zi2zi (Tian 2017) which modifies pixel2pixel (Isola et al 2017) to make it suitable for font generation. AGEN (Lyu et al 2017) proposes a model for synthesizing Chinese calligraphy images with specified style from standard font images.…”

Section: Many-shot Font Generationmentioning

confidence: 99%

VQ-FONT: Few-Shot Font Generation with Structure-Aware Enhancement and Quantization

Yao,

Zhang,

Lin

et al. 2024

AAAI

View full text Add to dashboard Cite

Few-shot font generation is challenging, as it needs to capture the fine-grained stroke styles from a limited set of reference glyphs, and then transfer to other characters, which are expected to have similar styles. However, due to the diversity and complexity of Chinese font styles, the synthesized glyphs of existing methods usually exhibit visible artifacts, such as missing details and distorted strokes. In this paper, we propose a VQGAN-based framework (i.e., VQ-Font) to enhance glyph fidelity through token prior refinement and structure-aware enhancement. Specifically, we pre-train a VQGAN to encapsulate font token prior within a code-book. Subsequently, VQ-Font refines the synthesized glyphs with the codebook to eliminate the domain gap between synthesized and real-world strokes. Furthermore, our VQ-Font leverages the inherent design of Chinese characters, where structure components such as radicals and character components are combined in specific arrangements, to recalibrate fine-grained styles based on references. This process improves the matching and fusion of styles at the structure level. Both modules collaborate to enhance the fidelity of the generated fonts. Experiments on a collected font dataset show that our VQ-Font outperforms the competing methods both quantitatively and qualitatively, especially in generating challenging styles. Our code is available at https://github.com/Yaomingshuai/VQ-Font.

show abstract

Section: Many-shot Font Generationmentioning

confidence: 99%

VQ-FONT: Few-Shot Font Generation with Structure-Aware Enhancement and Quantization

Yao,

Zhang,

Lin

et al. 2024

AAAI

View full text Add to dashboard Cite

show abstract

“…In the domain of image style transfer, there are two types of data inputs: paired character images, such as in pix2pix [6] and zi2zi [7], and unpaired character images, such as in CycleGAN [9] and CS-GAN [12].…”

Section: Related Workmentioning

confidence: 99%

“…However, this approach requires a significant amount of human intervention in generating Chinese characters and can often be complicated to implement. The second approach involves using image style transfer techniques to convert Chinese characters from one style to another [6,7,8,9]. This approach is fully automatic and requires no human intervention, making it relatively simple and effective.The existing methods such as pix2pix and zi2zi can generate fonts, although some of the fonts may be blurry.…”

Section: Introductionmentioning

confidence: 99%

A generative model for Chinese character style fusion based on generative adversarial networks

Zhou,

Dong,

Chen

2023

Third International Conference on Signal Image Processing and Communication (ICSIPC 2023)

View full text Add to dashboard Cite

Generative adversarial networks (GANs) have achieved remarkable success in image translation tasks. However, existing models and theories are still limited in their ability to generate new fonts that combine the styles of both font domains. Most current models only perform style conversion, which often results in generated fonts that are biased towards the target font domain, such as Chinese characters. Moreover, the long training time of these models makes them impractical for real-world applications. To address these issues, we propose a novel approach for font generation based on the pix2pix and AC-GAN frameworks. Our model can learn the mapping from the source font domain X to the target font domain Y without biasing the generated fonts towards either font domain. During training, the generator is optimized to produce high-quality new fonts that combine the styles of both font domains, while the discriminator is trained to distinguish between real and generated fonts. Experimental results show that our approach can generate 2555 new fonts with both font styles from a training set of 943 characters in a short time. The generated fonts are visually appealing and combine the styles of both font domains, achieving the goal of fusing styles and generating new fonts.

show abstract

“…Referring to image-to-image translation methods [2] between different domains, the HCCAG task is also regarded as the image-to-image style translation problem. In these related works, different font styles, such as DFKai-SB script 1 , running script, DFKai-SB and Pen-Kai scripts 2 , SIM-Kai script 3 and Lanting script 4 , are regarded as different data domains. Zi2zi [3] was the first work to use GAN to generate Chinese characters but required pairwise data as the input.…”

Section: Introductionmentioning

confidence: 99%

Radical Constraint-Based Generative Adversarial Network for Handwritten Chinese Character Generation

Ye,

Zhang,

Yang

et al. 2024

cai

View full text Add to dashboard Cite

Generative adversarial networks (GANs) have been used as a solution to handwritten Chinese character automatic generation (HCCAG) in recent years. However, most existing GAN-based methods adopt a pixel-based strategy, which ignores the radical structure of Chinese characters. To achieve better HCCAG, a radical constraint-based GAN (RC-GAN) is proposed in this work. In the proposed method, a gated recurrent unit (GRU)-based radical learning network is designed * Corresponding author Radical Constraint-Based GAN for HCCG 483 to calculate the radical components among Chinese characters, and radical consistent loss is applied to train this module. Finally, the radical learning module is fused with a cyclic structure GAN to improve the performance of Chinese character generation. The experimental results show that compared with the existing GAN, the proposed method has better performance. Specifically, the proposed method can reduce the stroke error in the generated Chinese character images.

show abstract

Colorization of Logo Sketch Based on Conditional Generative Adversarial Networks

Cited by 10 publications

References 33 publications

VQ-FONT: Few-Shot Font Generation with Structure-Aware Enhancement and Quantization

VQ-FONT: Few-Shot Font Generation with Structure-Aware Enhancement and Quantization

A generative model for Chinese character style fusion based on generative adversarial networks

Radical Constraint-Based Generative Adversarial Network for Handwritten Chinese Character Generation

Contact Info

Product

Resources

About