卷积神经网络的风格迁移算法在场景处理中的应用

The Application of Style Migration Algorithm of Convolution Neural Network in Scene Processing

针对传统风格迁移算法的诸多局限性,介绍了基于卷积神经网络的风格迁移算法,通过简化算法的生成步骤,实现图片的批量生成.就卷积神经网络计算过于耗时的问题,提出将像素损失改为感知损失,从而将运行时间降低了两个数量级.

Aiming at the limitations of traditional style migration algorithm,the style migration algorithm based on convolution neural network was introduced,which simplifies the generation step of the algorithm and realizes the batch generation of images.On the excessive time-consuming problem of neural network calculation,it is proposed to change the pixel loss to perceptual loss and reduce the running time by two orders of magnitude.