LPAdaIN: Light Progressive Attention Adaptive Instance Normalization Model for Style Transfer

To improve the generation quality of image style transfer, this paper proposes a light progressive attention adaptive instance normalization (LPAdaIN) model that combines (AdaIN) layer and convolutional block module (CBAM). In construction structure, first, lightweight autoencoder is built to reduce information loss in encoding process by reducing number network layers alleviate distortion stylized structure. Second, each AdaIN progressively applied after three relu encoder obtain fine-grained feature maps. Third, CBAM added between last decoder, ensuring main objects are clearly visible. optimization, reconstruction designed decoder’s ability decode images with more precise constraints refine structure images. Compared five classical transfer models, LPAdaIN visually shown finely apply texture content image, order which visible can be maintained. terms quantitative metrics, achieved good results running speed structural similarity.