Face swapping, a study of editing the face identity attributes, has received more attention in recent years due to the breakthrough progress of deep learning. Due to its application and technical value, this paper attempts to swap faces in a selected video clip. To begin with, we use the deep learning framework facenet pytorch to build a MTCNN model for face detection, next obtain the face feature vector based on FaceNet, and compare the Euclidean distance for target face recognition. Then use the InsightFace library to locate the 106 facial landmarks of the target person. Lastly, select the appropriate facial landmarks on the two-dimensional image to realize the pixel-based local translation operation, so that the face contour changes.

3D reconstruction is widely used in medical, film, game and other industries while point cloud data processing is the key to it. Considering the characteristics of the data obtained from athletes and sports students for 3D modeling, we describe a method to exact foot model containing down sampling with Open3D voxel as well as denoising with DBSCAN and radius filter algorithm. Compared with the mesh model generated from the front and rear foot point cloud data of denoising, it is proved that the method can effectively extract foot module data on the premise of removing a large number of noise points, which is convenient for the subsequent generation of insoles.

Film and television media are very commonly used in education and teaching. With the continuous development of the concept of "Internet & Education", the construction and application of a smart education platform featuring film and television is extremely important. This paper is based on the construction of a media-integrated learning platform “textbooks in film and television”, digital processing of textbook resources, and combined with data resources such as text, image, video, audio, 3D model and other data resources to build a knowledge base of course resources. Using Node.JS as the operating environment, the front-end uses the Vue3 framework, and the back-end uses Python & Flask technology to build a film and television characteristic smart education platform to help students master the content of knowledge points and related expansion, and improve their learning effects.

As a new method of film research, Cinemeitrics uses a systematic and digital way to measure and analyze film style, It focuses more on the measurability of objects and the accuracy of results. As an artistic medium, film has become a popular recording tool for creating historical documents. It is more and more actively getting rid of the shackles of text archives and playing an increasingly important role in recording the times, writing history. So this paper takes the historical film “The Founding of a Republic” as an example, uses image processing tools such as OpenCV to quantitatively analyze the film. The main work includes: shot segmentation, key frame extraction, subtitles extraction, object recognition and face tracking, providing data can be used to analyze film’s style, rhythm and emotion.

Nowadays, The network is developing rapidly and spreading all over the world, so a lot of data was generated, and we also felt the power of public opinion analysis. It can be said that every good or poorly-spoken movie will produce a wave of consensus for a period of time after release, and people are paying more and more attention to it. Our requirements for works and actors are becoming more and more strict. Therefore, whenever a good movie appears, the directors of various countries will remake, because there is already a mass foundation. This article mainly introduces the remake comparison between Chinese and foreign films and conducts a series of data visualization analysis, sorting, and comparison. In this paper, the data is visualized through analysis, and it is presented through three-dimensional word cloud, Nightingale rose diagram and force-Directed graph, so it is more convenient for observation. For the specific analysis of each film, this article has made a knowledge graph, which can be used to specifically view movies or characters.

In the thousands of years Chinese history, countless important figures have emerged. Some of their life experience, works and research results are data that can be processed by programs and visualized. China Biographical Database (CBDB) is a relational database on Chinese historical figures started by Harvard University. This paper using the above database, taking the literati of Song Dynasty as an example, constructed and visualized a social network based on the social relations and correspondence between the characters. PageRank algorithm is used to evaluate the importance of people in the social network.

Using the complex network theory to analyze the character social network in literatures provides a perspective in literary analysis. This paper proves that the character social networks of Harry Potter and The Lord of the Rings are small-world and scale-free and not only preferential attachment but also some specific plots affect the development of The Lord of the Rings trilogy. That means the social networks from two literatures obey the complex network theories that the social networks in real word follow. Hence, these above properties make the two fictional novels' networks more like networks in the real world and make people accept them well.

Entity relationship extraction is an important research topic in the field of NLP, which is to extract the relationship between two entities from natural language texts. Research shows that deep neural networks are feasible in the task of entity relation extraction and are superior to traditional relation extraction methods. In this paper, we compare the performance of the BLSTM model and the Bidirectional LSTM model with Attention under the entity relationship extraction task. The comparison includes the different performance of the model after adding additional NLP features and modifying embedding size. Compared with the BLSTM model, the Bidirectional LSTM model with Attention can achieve similar accuracy under the condition of simpler and less preprocessing. In addition, the larger size of dimension of word embedding and additional features acquired by NLP tools can improve the accuracy of relationship extraction. This article provides a reference standard for engineers to select the optimal model under different conditions.

Traditional text classification methods are based on statistics and feature selection. It does not perform well in processing large - scale corpus. In recent years, with the rapid development of deep learning and artificial neural networks, many scholars use them to solve text classification problems and achieve good results. Common text classification neural network models include textCNN, LSTM, and C-LSTM. Using a specific model can obtain more accurate features but ignore the context information. This paper proposes a C-LSTM with word embedding model to deal with this problem. Experiments show that the model proposed in this paper has great advantages in Chinese news text classification.

With the development of social networks, we are more interested in every aspect of the relationship. We conducted a visualization analysis of the papers published on two journals of Communication University of China in five years. In addition, we combine the academic social network of three art universities in Beijing based on CNKI and use complex network-related knowledge to analyze the collaboration of these universities. In this paper, we apply the knowledge of visualization tools and complex networks to analyze the authors who have high productivity, the keywords of these papers and the academic partnerships of the three universities. In experiments, each college is regarded as a social network model. From the perspective of academic social networks and collaborative networks, we extract some features of each of the three art universities and make some conclusions.

Nowadays, complex networks are widely used in various fields. As a hosting platform for open source and proprietary software projects, GitHub is used by many Internet companies and programmers. With the popularity of the Internet and the rapid development of the Internet, the number of users has exploded, and GitHub is not only a code hosting platform, but also a social networking platform for Internet practitioners to obtain information and knowledge, and even make friends. The sheer volume of users and the number of code repositories have also made it increasingly difficult for users to find code libraries that match their interests and users who share their interests. In order to effectively solve the above problems, based on the theory of ternary closure relations in the theory of social relations, according to the HITS algorithm, the repositories that may be interested and the like-minded friends are respectively recommended to the users. Another aspect of the system is data visualization, where users can intuitively get recommendations.

Assessing the importance of a node in a network is often very important. It can show the core of a network and provide guidance to decision makers. This paper combines the relevant characteristics of social networks, and proposes to use related smoothing weight factors to balance the clustering coefficient of nodes in the network, and to measure the importance of nodes as an important index. Experiments show that the importance of the nodes evaluated by this method is closer to the real situation. The importance of all nodes in the entire network is more consistent with the power law distribution.

The SALSA algorithm combines the main features of the PageRank and the HITS, which achieves better results than PageRank and HITS algorithm. The SALSA divides the nodes into two types of authority and hub as the HITS algorithm and adopts the random walk mode like the PageRank. Here tries applying the SALSA algorithm to the scene that recommends actors for films and TV dramas. The basic data is crawled by web crawler from douban.com, and it contains 500 films and 500 TV dramas with their chief four actors and scores. In this experiment, the films and TV dramas are seen as hub nodes, and the actors are seen as authority node. After hundred interactions of SALSA algorithm, a network index of each node is calculated. The higher the value is, the more worthy of recommendation.

In recent years, the network dramas have been more and more popular, and many people have become the screenwriter and director. How to create a better script has become a hot topic. This paper introduces the play script management system and discusses the realistic significance of character's biography and relationships for both screenwriters and directors. The system uses the online writing mode of the script to draw graphs of relationships and helps screenwriter and director better grasp the relationships among the characters and the plots of the story by visualization. Finally, the graph of relationships can not only help screenwriters create play scripts conveniently but also help directors understand dramas well.

This paper describes the importance of film and TV production projects management, the management solutions of TV network projects are provided by using contemporary computer technology. The shooting notice management is a sub module of the big data intelligent drama script system. It focuses on realizing automation and intelligent shooting notice management through improving the traditional way of writing and announcing shooting notices by using Excel spreadsheets. Be able to automatically supplement the other information of the field according to the completed field and avoid repeatedly finding the fill. Automatically query and fill out various personnel information. Determining the repeat of scene and the completion level of filming. Automatically judging expiration shooting notices. Archiving shooting notices for subsequent inspection. The announcement and management of the shooting notice is greatly facilitated.

With the continuous development of Internet technology, there is a growing desire for greater breakthroughs in the field of artificial intelligence to promote social development and progress. However, artificial intelligence has developed slowly in the past due to the complexity of its theory and the constraints of the related software and hardware conditions. As a branch of the field of AI, deep learning can learn the essential features of data and promote the development of AI. Convolution neural network is a typical multi-layer supervised learning neural network, which is widely used in various fields, especially image processing and speech recognition. This paper first introduces the research significance of convolution neural network, and then introduces its structure. The following paper studies and analyzes the architecture of LeNet-5, and improves it. Finally, the Keras framework is used to carry out the experiment on the improved network structure.

Nowadays, complex networks are widely used in various fields. We combine the role network of TV dramas and use complex network-related knowledge to analyze the characteristics of different types of popular TV series. In this paper, we apply the knowledge of social networks in a clever manner to analyze three different types of TV series-Mr. Love, West World and the Dream of Red Mansions. We introduce some important complex networks and their practical significance reflected in complex networks, including three-dimensional closures, structural holes, enemy-friend networks and the balance of multi-node network. In experiments, each TV series is abstracted as a social network model. From the perspective of social networks, we extract the above-mentioned feature structures in each network to find out their own characteristics and determine the appropriate ways to measure them.

The present paper concerns itself with a theoretical calculation study on implementation and optimization of protein sequence issue. Using ant colony optimization algorithm (ACO) to build the Two-Dimensional Hydrophobic-Hydrophilic (2DHP) model. There are three parts in this paper. Starting with a brief introduction of the ant colony optimization and 2DHP model, then we propose a method on how to construct such model based on ant colony optimization algorithm. We end the study with the conclusion that ant colony optimization algorithm for protein sequence theoretical calculation can be improved.

This paper mainly talks about the Recurrent Neural Network and introduces a more effective neural network model named LSTM. Then, the paper recommends a special language model based on Recurrent Neural Network. With the help of LSTM and RNN language models, program can predict the next character after a certain character. The main purpose of this paper is to compare the LSTM model with the standard RNN model and see their results in character prediction. So we can see the huge potential of Recurrent Neural Network Language Model in the field of character prediction.

with the continuous development of Internet technology, information overload is becoming more and more serious. It's getting harder to get useful information from the network. Although the search engine can help users find information they need from the vast amounts of information in a certain extent, but cannot completely solve the problem of information overload, when users cannot accurately describe the information they need, you need to recommend system to help users find valuable information for users. So recommender systems are becoming more and more important. The movie recommender system implemented in this paper is based on the traditional user-based collaborative filtering algorithm, and the user project scoring matrix is pre filled. At the same time, database technology of this system uses graph database which is good at dealing with complex relations. In data visualization, the degree of recommendation of a movie is expressed by the size of the node and the thickness of the edge, so as to improve the user experience.

Unity 3D is a cross-platform 3D game engine. That is an important problem of seeking the road in the process of game production. The main solution is path-finding algorithm in which the player finds a path to the target in game scene. In this paper, by general path-finding game scene as background, we focus on the A* algorithm and “Navmesh Grid” path-finding in Unity 3D, and put forward the improvement and application of algorithm in the special game scene.

This paper propose a new method to extract ontology concepts from multiple text of the same type. This method uses mutual information and document frequency. According to the mutual information tend to choose the low frequency words, combining mutual information and document frequency to avoid this problem. In this paper, a number of financial and economic reports are reported as the corpus. First of all, do the text preprocessing, and then based on the N-gram algorithm to generate a set of candidate phrases, and finally use the statistics and the rules to screen for the concept of ontology from candidate phrases.

With the development of the Web, social media, the SINA Microblog for instance, is gaining wider popularity. This indicates most users' attitudes towards emerging technologies that happened recently. To obtain the data on social media, the traditional Web crawler is only able to obtain part of the information by collecting data in a non-login status because the crawler is not provided with the capabilities to log into the system. To enable information sharing among users, the SINA Microblog Open Platform offers a number APIs in which certain limitations such as those on the number of user requests of the microblog server, still exist. In this paper, a method that simulates user login to collect Microblog data related to hot topics on the SINA platform is proposed. What's more, the method overcomes the limitations of the SINA Microblog APIs, thus leading to getting more data from the system. In addition, the typical user behaviors, through the collected data, is found when they interact with social media.

Analyzed here is a fault localization approach based on directed graph in view point of business software. The fault propagation model solved the problem of obtaining the dependency relationship of faults and symptoms semi-automatically. The main idea includes: get the deployment graph of managed business from the topography of network and software environment; generate the adjacency matrix of the graph; compute the transitive closure of adjacency matrix and obtain a so-called dependency matrix; independent fault locates in the main diagonals of the dependency matrix, elements of column is fault’s symptoms domain, which are divided into immediate effect and transitive effect. In real world, those elements will denote symptoms, but the two class affected nodes have different probability of occurrence, that is, immediate symptoms can be observed more likely than transitive symptoms and symptom of fault itself is most likely observed. Based on the hypothesis, a new fault localization algorithm is proposed. The simulation results show the validity and efficiency of this fault localization approach.