Creative Design of Motion Graphics Based on BicycleGAN Algorithm
With the growth of computer technology and graphic design, to solve the pattern collapse of the generator and improve the diversity of generated results, a study was conducted and designed based on BicycleGAN. A Map2Style module was added to modify the encoder of the model to obtain encoding feature...
Main Authors: | , |
---|---|
Format: | Article |
Language: | English |
Published: |
Elsevier
2024-03-01
|
Series: | Intelligent Systems with Applications |
Subjects: | |
Online Access: | http://www.sciencedirect.com/science/article/pii/S2667305323001485 |
_version_ | 1797287551935971328 |
---|---|
author | Jie Lin Rui Li |
author_facet | Jie Lin Rui Li |
author_sort | Jie Lin |
collection | DOAJ |
description | With the growth of computer technology and graphic design, to solve the pattern collapse of the generator and improve the diversity of generated results, a study was conducted and designed based on BicycleGAN. A Map2Style module was added to modify the encoder of the model to obtain encoding feature information, thereby improving the accuracy of the encoder and solving the pattern collapse of the generator, thereby increasing the diversity of generated results. The results denoted that the structural similarity indicators of sapphire and white effects were both 0.93. The structural similarity index of the water effect was 0.59, and the peak signal-to-noise ratio index was 28.16. When the batch size was the same, the training time was the same for pattern search generative adversarial network, Bicycle generative adversarial network, and improved method. At this point, the three occupied 3.6, 3.2, and 3.3 GB of video memory, respectively. The area between the receiver operating characteristic curve of the improved model and the coordinate axis was 0.9837, which was higher than that of the Bicycle generative adversarial network, denoting that the improved model had more accurate effectiveness. The performance accuracy of the improved model was 96.31%, and the sensitivity was 80.65%. The improved model has a significant effect, making it more stable and achieving convergence faster during training. It plays a certain role in promoting the application of motion graphics creative design. |
first_indexed | 2024-03-07T18:34:50Z |
format | Article |
id | doaj.art-30024213cebe42dd9d09fb6e00c0b285 |
institution | Directory Open Access Journal |
issn | 2667-3053 |
language | English |
last_indexed | 2024-03-07T18:34:50Z |
publishDate | 2024-03-01 |
publisher | Elsevier |
record_format | Article |
series | Intelligent Systems with Applications |
spelling | doaj.art-30024213cebe42dd9d09fb6e00c0b2852024-03-02T04:55:19ZengElsevierIntelligent Systems with Applications2667-30532024-03-0121200323Creative Design of Motion Graphics Based on BicycleGAN AlgorithmJie Lin0Rui Li1School of art and design, Xiangsihu College of Guangxi Minzu University, Nanning, 530225, ChinaCorresponding author; School of art and design, Xiangsihu College of Guangxi Minzu University, Nanning, 530225, ChinaWith the growth of computer technology and graphic design, to solve the pattern collapse of the generator and improve the diversity of generated results, a study was conducted and designed based on BicycleGAN. A Map2Style module was added to modify the encoder of the model to obtain encoding feature information, thereby improving the accuracy of the encoder and solving the pattern collapse of the generator, thereby increasing the diversity of generated results. The results denoted that the structural similarity indicators of sapphire and white effects were both 0.93. The structural similarity index of the water effect was 0.59, and the peak signal-to-noise ratio index was 28.16. When the batch size was the same, the training time was the same for pattern search generative adversarial network, Bicycle generative adversarial network, and improved method. At this point, the three occupied 3.6, 3.2, and 3.3 GB of video memory, respectively. The area between the receiver operating characteristic curve of the improved model and the coordinate axis was 0.9837, which was higher than that of the Bicycle generative adversarial network, denoting that the improved model had more accurate effectiveness. The performance accuracy of the improved model was 96.31%, and the sensitivity was 80.65%. The improved model has a significant effect, making it more stable and achieving convergence faster during training. It plays a certain role in promoting the application of motion graphics creative design.http://www.sciencedirect.com/science/article/pii/S2667305323001485BicycleGAN algorithmMotion graphicsCreative designGeneratorEncoder |
spellingShingle | Jie Lin Rui Li Creative Design of Motion Graphics Based on BicycleGAN Algorithm Intelligent Systems with Applications BicycleGAN algorithm Motion graphics Creative design Generator Encoder |
title | Creative Design of Motion Graphics Based on BicycleGAN Algorithm |
title_full | Creative Design of Motion Graphics Based on BicycleGAN Algorithm |
title_fullStr | Creative Design of Motion Graphics Based on BicycleGAN Algorithm |
title_full_unstemmed | Creative Design of Motion Graphics Based on BicycleGAN Algorithm |
title_short | Creative Design of Motion Graphics Based on BicycleGAN Algorithm |
title_sort | creative design of motion graphics based on bicyclegan algorithm |
topic | BicycleGAN algorithm Motion graphics Creative design Generator Encoder |
url | http://www.sciencedirect.com/science/article/pii/S2667305323001485 |
work_keys_str_mv | AT jielin creativedesignofmotiongraphicsbasedonbicycleganalgorithm AT ruili creativedesignofmotiongraphicsbasedonbicycleganalgorithm |