基于VGG-16神经网络图像风格迁移模型

Image Style Transfer Model Based on VGG-16 Neural Network

图像风格迁移是人工智能进行艺术创造的一个重要方向。传统风格迁移技术通过逐像素迭代得到风格图片,训练耗时且迁移效果一般,无法广泛地应用于微端设备上。针对此问题,本文提出了一款轻量的图像风格迁移模型,该模型能够充分利用VGG-16卷积网络强大的图像特征提取功能。通过优化兼顾了图像内容和风格信息的损失函数,该模型能够在短时间内完成图像的风格学习,并迁移运用到目标图片上,所得到的迁移图片效果优于传统风格迁移技术。

Image style transfer is an important direction of art creation by artificial intelligence.The traditional style transfer technology gets the style image by pixel iteration,which is time-consuming to train and the transfer effect is not good,so it can't be widely applied to the micro-terminal device.To solve this problem,this paper proposes a lightweight image style transfer model,which could make full use of the powerful image feature extraction function of VGG-16 convolutional network.By optimizing the loss function of image content and style information,the model can complete image style learning in a short time,and transfer to the target image,and the obtained transfer image effect is better than the traditional style transfer technology.