This study aims to deeply study the training and optimization methods of image style transfer models based on CoreML technology, hoping to achieve higher-quality image style conversion. In this study, different training algorithms, hyperparameter adjustment, and data enhancement techniques are studied to optimize the performance of the model. The experimental results reveal that the U-Net style transfer model shows the best style transfer metric. Therefore, this study provides not only a more powerful tool for image processing applications on mobile devices but also new ideas for further research in the field of image style transfer.

Xuanhui You. Research on Training and Optimization of Image Style Transfer Model Based on CoreML. Academic Journal of Computing & Information Science (2023), Vol. 6, Issue 13: 47-51. https://doi.org/10.25236/AJCIS.2023.061307.

[1] Magotra A , Kim J .Improvement of Heterogeneous Transfer Learning Efficiency by Using Hebbian Learning Principle[J].Applied Sciences, 2020, 10(16):5631.

[2] Shi Yongchao; Zhu Lijun. Research on Image Style Transfer Based on GAN [J]. Electronic Technology and Software Engineering, 2020:4.

[3] Hong,Choe.Wasserstein Generative Adversarial Network Based De-Blurring Using Perceptual Similarity[J].Applied Sciences, 2019, 9(11):2358.

[4] Yang Bo; Zhou Yicheng. Design and Implementation of Image Style Transfer Based on Generative Models [J]. Journal of Hunan University of Technology (Natural Science Edition), 2020:6.

[5] Wu K , Qiang Y , Song K ,et al.Image synthesis in contrast MRI based on super resolution reconstruction with multi-refinement cycle-consistent generative adversarial networks[J].Journal of Intelligent Manufacturing, 2020, 31.

[6] Zhao C , Zhao H , Wang G ,et al.Hybrid Depth-Separable Residual Networks for Hyperspectral Image Classification[J].Complexity, 2020.

[7] Zhong K, Wang Y, Pei J, et al. Super efficiency SBM-DEA and neural network for performance evaluation[J]. Information Processing & Management, 2021, 58(6): 102728.