ABSTRACT
The main task of the font design is to design a suitable font according to the actual application scenario, which has extremely wide commercial application value. Generally the traditional font design requires professionals to design, with longer design time, lower work efficiency, and higher labor costs. Font design is essentially a problem of image synthesis. U-net is a deep learning network structure, which has been widely used in image synthesis, but the images synthesized by U-net have the disadvantages of low image quality and poor visual effects. In order to improve the shortcomings of U-net image synthesis effectively, this paper provides an improved U-net method for better font design. The new method is called Swish-gated residual dilated U-net (Swish-gated residual dilated U-net, SGRDU). In SGRDU, the proposed swish layer and swish-gated residual block can effectively control the information transmitted by each horizontal and vertical layer in U-net, and accelerate the network convergence. Dilated convolution is used to improve the perception of the network. Experimental results show that, compared with other residual U-net, the font synthesized by SGRDU has better visual effect and quality.
- Schmidhuber, J.: Deep learning in neural networks: an overview. Neural Netw. 61, 85–117 (2015)Google ScholarDigital Library
- Liang Peijun, Liu Yijun. Coloring method of hand-drawn cartoons based on conditional generation of confrontation networks [J]. Computer Application Research, 2019.36(01): 308-311.Google Scholar
- Wu Xiaoqi. Research on manga style transfer method based on deep learning [D]. Xi'an University of Technology, 2019.Google Scholar
- Ronneberger, O., Fischer, P., Brox, T.: U-Net: convolutional networks for biomedical image segmentation. In: Navab, N., Hornegger, J., Wells, W.M., Frangi, A.F. (eds.) MICCAI 2015. LNCS, vol. 9351, pp. 234–241. Springer, Cham (2015).Google Scholar
- Dong, H., Yang, G., Liu, F., Mo, Y., Guo, Y.: Automatic brain tumor detection and segmentation using U-Net based fully convolutional networks. In: Vald´esHern´andez, M., Gonz´alez-Castro, V. (eds.) MIUA 2017. CCIS, vol. 723, pp. 506– 517. Springer, Cham (2017).Google Scholar
- Liu G., Chen X., Hu Y. (2019) Anime Sketch Coloring with Swish-Gated Residual U-Net. In: Peng H., Deng C., Wu Z., Liu Y. (eds) Computational Intelligence and Intelligent Systems. ISICA 2018. Communications in Computer and Information Science, vol 986. Springer, SingaporeGoogle Scholar
- Fisher Y.Vladlen K.:Multi-Scale Context Aggregation by Dilated Convolutions .arXiv:1511.07122[cs.CV](2016)Google Scholar
- Johnson, J., Alahi, A., Fei-Fei, L.: Perceptual losses for real-time style transfer and super-resolution. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds.) ECCV 2016. LNCS, vol. 9906, pp. 694–711. Springer, Cham (2016)Google Scholar
- .Lin, B.S., Michael, K., Kalra, S., Tizhoosh, H.R.: Skin lesion segmentation: U-nets versus clustering. In: Proceedings of 2017 IEEE Symposium Series on Computational Intelligence (SSCI 2017), Honolulu, HI, United States, pp. 1–7, November 2017Google ScholarCross Ref
- Zhao, H., Sun, N.: Improved U-net model for nerve segmentation. In: Zhao, Y., Kong, X., Taubman, D. (eds.) ICIG 2017. LNCS, vol. 10667, pp. 496–504. Springer, Cham (2017).Google Scholar
- He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of 29th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2016, Las Vegas, NV, United states, pp. 770–778 (2016)Google ScholarCross Ref
- Ramachandran, P., Zoph, B., Le, Q.V.: Searching for activation functions. CoRR abs/1710.05941 (2017).Google Scholar
- He, K., Zhang, X., Ren, S., Sun, J.: Delving deep into rectififiers: surpassing humanlevel performance on imagenet classifification. In: Proceedings of 2015 IEEE International Conference on Computer Vision (ICCV 2015), Santiago, Chile, pp. 1026– 1034 (2015)Google ScholarDigital Library
- Ba, J.L., Kiros, J.R., Hinton, G.E.: Layer normalization. CoRR abs/1607.06450 (2016).Google Scholar
- Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. CoRR abs/1409.1556 (2014).Google Scholar
- Zhang, Z., Liu, Q., Wang, Y.: Road extraction by deep residual U-net. IEEE Geosci. Remote Sens. Lett. 15(5), 749–753 (2018)Google ScholarCross Ref
- Huang Jiaheng, Li Xiaowei, Chen Benhui, Yang Dengqi. A comparative study of image similarity algorithms based on hash [J]. Journal of Dali University, 2017, 2(12): 32-37.Google Scholar
- Wang, Z., Bovik, A.C., Sheikh, H.R., Simoncelli, E.P.: Image quality assessment: from error visibility to structural similarity. IEEE Trans. Image Process. 13(4), 600–612 (2004)Google ScholarDigital Library
Recommendations
MF-Net: A Novel Few-shot Stylized Multilingual Font Generation Method
MM '22: Proceedings of the 30th ACM International Conference on MultimediaCreating a complete stylized font library that helps the audience to perceive information from the text often requires years of study and proficiency in the use of many professional tools. Accordingly, automatic stylized font generation in a deep ...
A U-Net based Self-Supervised Image Generation Model Applying PCA using Small Datasets
CACML '23: Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine LearningGenerative Adversarial Networks (GAN) is a research-based on deep learning technology that synthetically generates, combines, and transforms images similar to the original images. The main focus of GAN existing work has been to improve the quality of ...
Projection-Based cascaded U-Net model for MR image reconstruction
Highlights- MR image reconstruction is an important inverse problem in medical imaging.
- ...
Abstract Background and ObjectiveBackground and Objective: Recent studies in deep learning reveal that the U-Net stands out among the diverse set of deep models as an effective network structure, especially for imaging inverse ...
Comments