Fast Continuous Structural Similarity Patch Based Arbitrary Style Transfer

Abstract

Style transfer is using a pair of content and style images to synthesize a stylized image which has both the structure of the content image and the style of style image. Existing optimization-based methods are limited in their performance. Some works using a feed-forward network allow arbitrary style transfer but cannot reflect the style. In this paper, we present a fast continuous structural similarity patch based arbitrary style transfer. Firstly, we introduce the structural similarity index (SSIM) to compute the similarity between all of the content and style patches for obtaining their similarity. Then a local style patch choosing procedure is applied to maximize the utilization of all style patches and make the swapped style patch continuous matching with respect to the spatial location of style at the same time. Finally, we apply an efficient trained feed-forward inverse network to obtain the final stylized image. We use more than 80,000 natural images and 120,000 style images to train that feed-forward inverse network. The results show that our method is able to transfer arbitrary style with consistency, and the result comparison stage is made to show the effectiveness and high-quality of our stylized images.

HaCk: Hand Gesture Classification Using a Convolutional Neural Network and Generative Adversarial Network-Based Data Generation Model

“…One popular approach involves manipulating the intermediate features of a CNN. The gram matrix was utilized for CNN's middle feature extraction from a pretrained VGG network to represent style [20][21][22][23][24][25][26][27][28]. Similarly, Holden et al proposed analogous ideas for motion stylization [29].…”

“…Similarly, Holden et al proposed analogous ideas for motion stylization [29]. Wu et al proposed patch swapping between feature maps of the input and style images to achieve stylization [20]. Sheng et al proposed the features of whitening and de-whitening to preserve the input image's global structure while reflecting the reference image's style [27].…”

Research on digital restoration and redesign of Taishan stone carving

“…Additionally, considerable attention has been directed towards threedimensional modeling technology [6]. For instance, Cha Hongbin and colleagues have implemented three-dimensional laser scanning and photography as primary methodologies for data acquisition [7]. These scholarly endeavors underscore China's dedication to innovative methods to conserve and comprehensively study the cultural heritage embodied in stone carvings.…”

Research on big data women's portrait platform based on knowledge graph analysis

“…Associate Professor Bai of Anhui Xinhua College and others entirely use network teaching resources to reform the online and offline hybrid teaching mode of the "Internet of Things Application Technology" course. Prompting the effective connection between online and offline teaching enhances the students' learning autonomy [4], improves the effectiveness and quality of teaching, and is well received by the students.…”

“…Knowledge mapping is called knowledge domain visualization or knowledge domain mapping map in the library and intelligence community [9], which is a series of different graphics showing the process of knowledge development and structural relationships, describing the knowledge resources and their carriers with visualization technology, mining, analyzing, constructing, drawing and displaying the knowledge and the interconnection between them [10]. Applying it in personalized platforms research based on learner profiling is conducive to a more intuitive presentation of learner information [11].…”