<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>Image style transfer is a major area of study in image processing and has applications in creative production, special effects for film and television, and other areas. Image style transfer is the process of using style transfer technology to change a common image into one with a different style without changing the content. Image style transfer methods are mainly divided into traditional image style transfer methods and deep learning image style transfer methods. The two primary classifications of picture style transfer techniques are deep learning technologies and conventional methods. Traditional image style transfer methods have poor results and are difficult to apply in people's lives. With the quick advancements in machine learning, digital image processing, and computer vision, deep learning image style transfer methods have received widespread attention from researchers. Most of these methods use convolutional neural networks to achieve image style transfer on the premise of paired data sets, but obtaining paired data sets is difficult and costly. Accordingly, it is of great significance to study unpaired images to implement style transfer algorithms. The primary focus of this study is the CycleGAN network-based picture style transfer technique, and improves this algorithm in content compiler, style compiler. It is applied to the generation of night road conditions during autonomous driving training.</p>
</abstract>

Image style transfer is a major area of study in image processing and has applications in creative production, special effects for film and television, and other areas. Image style transfer is the process of using style transfer technology to change a common image into one with a different style without changing the content. Image style transfer methods are mainly divided into traditional image style transfer methods and deep learning image style transfer methods. The two primary classifications of picture style transfer techniques are deep learning technologies and conventional methods. Traditional image style transfer methods have poor results and are difficult to apply in people's lives. With the quick advancements in machine learning, digital image processing, and computer vision, deep learning image style transfer methods have received widespread attention from researchers. Most of these methods use convolutional neural networks to achieve image style transfer on the premise of paired data sets, but obtaining paired data sets is difficult and costly. Accordingly, it is of great significance to study unpaired images to implement style transfer algorithms. The primary focus of this study is the CycleGAN network-based picture style transfer technique, and improves this algorithm in content compiler, style compiler. It is applied to the generation of night road conditions during autonomous driving training.

Research on the Expanded Night Road Condition Dataset Based on the Improved CycleGAN

School of Computer Science and Engineering

International Journal of Advanced Network, Monitoring and Controls

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Image style transfer is a major area of study in image processing and has applications in creative production, special effects for film and television, and...

Research on the Expanded Night Road Condition Dataset Based on the Improved CycleGAN

Publié en ligne: 21 juil. 2024

Pages: 59 - 66

DOI: https://doi.org/10.2478/ijanmc-2024-0016

Mots clés
Generative, Adversarial Networks, Style Migration, Deep Learning

© 2024 Lei Cao et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Research on the Expanded Night Road Condition Dataset Based on the Improved CycleGAN

Publié en ligne: 21 juil. 2024

Pages: 59 - 66

DOI: https://doi.org/10.2478/ijanmc-2024-0016

Mots clésGenerative, Adversarial Networks, Style Migration, Deep Learning

© 2024 Lei Cao et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Mots clés
Generative, Adversarial Networks, Style Migration, Deep Learning