Application of big data and artificial intelligence in visual communication art design

Application of AI in VISCOM

There are many feasible examples of the application of AI in VISCOM art design, such as identification and classification in text design, 3D printing, automated packaging, etc. In this essay, the application of AI technology is mainly studied from the main design contents of VISCOM art design. The main contents of VISCOM art design include scene design, poster design, logo design, packaging design, and text design (Rehman, Po & Liu, 2019; Constant, 2018). Standard AI technologies for text design include text recognition and intelligent text reading. For scene design, there are 3D scene intelligent construction, virtual scene interaction, etc. Packaging design includes automatic packaging, automatic classification of packaging, etc., as shown in Fig. 1.

Many companies even pay special attention to the application and research of AI. To respond to the national call and keep up with the pulse of development, Baidu, Tencent, Ali, and other companies have actively invested in technical research and AI application practice. As the company with the largest investment, the most extensive distribution and the most robust overall strength in the domestic AI field, Baidu established the Baidu Silicon Valley AI Lab very early. Tencent established AILab in 2016. It spent a lot of money to recruit outstanding R&D personnel in the AI field, intending to accelerate the process of AI. Tencent’s cloud platform also provides standard application technologies such as image, speech, and natural language processing. Alibaba mainly focused on its own e-commerce business and commercial fields and has built the Alibaba Cloud Vision Platform.

At the same time, there has also been a significant increase in the number of AI-related patent applications, patent grants, and related papers published in China, proving that domestic AI research is constantly being explored and applied (Liu et al., 2018; Zhu et al., 2017). The development of AI can begin with a test in 1950 that boldly predicted the possibility of machines with human intelligence. To verify whether the machine has intelligence, it has been proposed that the human and the machine accept the question of the observer. If the observer cannot distinguish the difference between the two, the machine is considered to have intelligence, which is the famous Turing test. After the emergence of neural network algorithms, the research focus of AI has also expanded in the fields of expert systems, language translation, image interpretation, pattern recognition, fault diagnosis, and intelligent control. After the 1990s, computer performance improved; further, accumulating a large amount of data and researchers’ research made AI breakthroughs in many fields, and considerable practical results were continuously achieved (Zhu, Luo & Dai, 2018; Li, Wen & Kuai, 2019). With the information expansion and the rapid development of technology, the whole industry’s market demand and operation tools are constantly updated and iterated. To maintain the value of art practitioners, the only way to adapt to the changing work mode is to use AI technology to improve work efficiency. In the future, the art design market will present a new situation of collaborative innovation and cooperation between AI and human brain intelligence. The demand in the art and design industry also reflects that researchers can no longer limit themselves to their specialties but must seek new paths that meet the development needs of the times by integrating multiple disciplines. It is necessary to advocate multi-disciplinary crossover, break through the boundaries of disciplines, and build a new integration. The core competitive factors are art practitioners’ creativity, forward-lookingness, and self-value expression. The perspective of AI intervention in visual art is to promote multi-disciplinary cross-learning, which provides people with ways to integrate inter-professionals. It is easier for researchers to cut into other fields or for scholars in different fields to enter the field of design art (Foumani & Nickabadi, 2019; Xie, Wu & Yang, 2019).

Application of big data in VISCOM

The rapid development of information technology inevitably generates enormous amounts of data. This explosive amount of data has wholly exceeded people’s imagination. Only a tiny amount of such massive data is stored in media such as newspapers and books, and most of them are stored in the form of digital data. Therefore, people’s lives have also undergone tremendous changes. The current government, medical and health care, public safety, and urban construction all use big data to make predictions. This data is like the value of “petroleum”. Therefore, it is essential to display the data information using VISCOM (Hu, Lam & Lou, 2018). The emergence of big data significantly influences human culture and continually influences both sections of conventional artistic production. In the last ten years, creators have begun to listen closely to statistics as a motivator since it is the sign of software applications or the feedstock in the age of big data. These data permeate every corner and affect the artist’s thinking and creation (Morra, Lamberti & Pratticó, 2019; Lefebvre, Chen & Beauseroy, 2017). The related technologies of big data are used to fully understand the characteristics of VISCOM art creation under the background of new technologies. A new way of visual art expression assisted by big data technology is studied, which expands creativity and enhances artists’ spontaneous integration. In this essay, the application of VISCOM art design through big data-related technologies is also carried out according to its main design content, as shown in Fig. 2.

In packaging design, big data can be used to classify packaging products. In the poster design, information mining can be carried out on the location or scene of the poster to design a more suitable poster. In logo design, the copyright can be quickly obtained through big data to reduce the risk. In the scene design, the scene can be distinguished by collecting information about the scene. In text design, big data can classify text and personalize text information. Using these big data methods, the effect of visual transmission can be implemented more intelligently and efficiently (Lin & Zhao, 2020; Tang, Gao & Lin, 2018). Today, people have entered a new information society, and the leap from data to big data is also due to the advent of the Internet age. It has increasingly become a strategic proposal for today’s society. The Internet and mass media development has brought big data and VISCOM art into people’s daily lives. VISCOM uses a unique information design method to produce visually impactful works of art. Through comprehensive disciplinary research, the data and information can be effectively transmitted, and the information can be more shining (Zhang, Liu & Hao, 2018; Li, Peng & Nai, 2018). It not only possesses scientific rigor but also endows art with aesthetic value. Its unique artistic creation method transforms difficult and tedious data into visual works with aesthetic effect.

STING algorithm

For the complex needs of VISCOM art design, the powerful multi-resolution clustering technology known as STING was chosen due to its outstanding efficiency in handling big datasets. STING uses a hierarchical grid layout to partition the data space to facilitate quick and scalable data clustering, with each cell representing a distinct area. The algorithm precomputes and maintains critical statistical metrics like the standard deviation and mean for every cell to facilitate fast query answers and precise clustering without the usual computing load. The capacity to examine complicated, multi-dimensional data at different degrees of detail is vital for VISCOM, which is why this grid-based method benefits them. Our study uses STING to identify design elements while drastically reducing data processing time precisely. This optimizes the clustering process in terms of accuracy and efficiency. As a multi-resolution clustering technology for big data, STING has the advantage of efficient data processing. This essay uses the algorithm to cluster and analyze the object information in VISCOM art design to mine and summarize the hidden elements and attributes in the design objects. The algorithm uses the number of objects in the count grid and the average of all values in the mean grid as parameters commonly used in grids. Statistics for each grid cell attribute (such as mean, maximum, and minimum) are precomputed and stored. These statistics are used to respond to queries. Compared with other clustering algorithms, STING has the advantage of faster data processing (Zheng, Yu & Huang, 2018). The first step of the STING query algorithm is to build an object hierarchy. Then, a query level is selected, where it starts to analyze the object properties. Then, the next level of query is carried out. If this layer is the bottom layer, the query result is satisfied; otherwise, go to the next layer query according to the hierarchy. Finally, the query is stopped after the query results are satisfied. The query process is relatively simple, so the data clustering processing speed is very fast (Sun et al., 2020; Hu et al., 2021).

CNN recognition algorithm

CNN is a very widely used intelligent algorithm. It can also perform feature extraction efficiently and quickly, which is especially common in AI. The CNN architecture for scene and facial recognition combines accuracy and computational efficiency, making it suitable for VISCOM art creation. It comprises five convolutional layers with a 3x3 filter size and ReLU activation function. Max-pooling layers reduce feature map size, allowing the network to acquire hierarchical characteristics. The three fully linked layers integrate these characteristics, generating scene and facial recognition predictions. The softmax activation function aids in multi-class classification. This design ensures efficient processing of intricate VISCOM art data without significant speed decrease. This essay uses the CNN algorithm to intelligently identify works in VISCOM art design. At the same time, it can be identified and analyzed in the poster placement scene to match the more suitable VISCOM art. In addition, the CNN algorithm can be used to perform face recognition and combined with big data to deliver VISCOM artworks that are ideal for different groups of people. Regarding the identification process of CNN, the details are as follows:

It is assumed that the design object in VISCOM art design contains m design elements, and the CNN model has L layers. The size of the convolution kernel is k, the weight between each layer is represented by W, and b is the adjustment coefficient. Then, the convolution process is usually: (1)x j l = f ∑ x i l − 1 ∗ k i j l + b j l

During backpropagation, the error of the output layer is first calculated: (2)δ l = ∂ J W , b ∂ z L = ∂ J W , b ∂ α L ⊗ σ ′ z L

The error of the previous layer can be calculated using recursion: (3)δ j l = α i l + 1 f ′ u l − u δ j l + 1

Among them, u represents the pooling layer. The overall error can be calculated as: (4)∂ E ∂ L = ∑ i δ i l

After knowing the error gradient, the weight expression can be calculated: (5)∂ J W , b ∂ W l = δ l α l − 1 T

In the forward propagation process, the data matrix α can be expressed as: (6)α l = σ z l = σ α l − 1 ∗ W l + b l

According to the error formula, it can be obtained: (7)δ l − 1 = ∂ z l ∂ z l − 1 T δ l = δ l ∗ r o t 180 W l ⊗ σ ′ z l − 1

The update method of the weight value and the update method of the adjustment coefficient are: (8)W l = W l − α ∑ i = 1 m α i , l − 1 ∗ δ i , l (9)b l = b l − α ∑ i = 1 m ∑ u , v δ i , l u , v

From this, the error of each layer is iteratively calculated, and finally, classification and identification can be performed according to the error.

Big data clustering results

Regarding the big data clustering method used in this essay, the number of results and accuracy of the clustering are shown in Fig. 3. As seen from the figure, in the logo design of VISCOM art, there were about 950 elements related to the design objects clustered in this essay. In the poster design, there were more than 680 related elements. In the scene design, there were about 1,200 related elements, about 335 text design elements, about 515 packaging design elements, and an overall average of about 735 elements, and this shows that the clustering algorithm in this essay can provide many design elements for various WISCOM art designs, which greatly facilitates the design.

In addition, among the clustering accuracy of these elements, the accuracy of logo design and text design was relatively high, about 90%, about 84% for scene design, and about 79% for packaging design. The accuracy of element clustering for poster design was only about 68%. This may be because the element information in the poster design is relatively complex, and the error generated in the clustering process is large. However, the overall average clustering accuracy was above 82%, which showed that this essay used the big data STING algorithm to cluster VISCOM art design with high accuracy and had certain practicability. Aside from the accuracy data already provided, we conducted additional assessments of the CNN and STING algorithms using precision, recall, and F1 score measures. These metrics offer a more detailed comprehension of the algorithms’ capabilities. The accuracy, which quantifies the ratio of accurate positive predictions to all positive predictions, has an average value of approximately 85% across all design tasks, with facial recognition exhibiting notably high precision. The recall, which measures the algorithm’s capacity to correctly identify all pertinent instances in the dataset, was approximately 82%. This indicates that the algorithm is effective at detecting design aspects, although its performance may vary depending on the complexity of the scenario. The F1 score, calculated as the harmonic mean of precision and recall, has an average value of 83%. This value indicates a high overall performance in achieving a balance between accuracy and sensitivity. In addition to accuracy, these supplementary metrics thoroughly assess the algorithms’ efficacy in VISCOM art design.

AI recognition effect

Next, the AI technology in this essay is used to identify the layout scenes and viewing groups of VISCOM artworks. The number and accuracy of the 16 design items identified are summarized in Fig. 4.

Due to the different scenes, the number of elements recognized by the CNN algorithm in scene recognition was different. There were at least four in the picture and 29 at most. The difference was quite big. Generally speaking, a relatively single scene can be recognized with fewer elements, and a more complex scene can be recognized with more elements. In crowd facial recognition, there were differences in the number of people recognized due to different population flow. There were only 211 less in the picture and 534 more. In scene recognition, the accuracy of scene element recognition was between 73% and 89%, most of which were above 80%. The accuracy of facial recognition was more than 80%. The highest was more than 94%, which shows that the use of the CNN algorithm in AI technology for facial recognition in this essay has certain reliability. This study assessed the dependability and feasibility of artificial intelligence and big data methods in VISCOM art design. The evaluation criteria encompassed the factors of accuracy, efficiency, inventiveness, and usability. The procedure consisted of several steps in which evaluators were given comprehensive information and independently reviewed design outcomes. The panel had five professionals who possessed more than a decade of experience in visual communication, graphic design, and the utilization of AI in art. The meticulous method of evaluation ensured the reliability of the scores.

Overall rating

Finally, this essay evaluates the application effect of big data and AI in terms of reliability and practicability through expert scoring. The evaluation results are shown in Fig. 5.

As can be seen from Fig. 5A, the application reliability scores of big data in various designs of VISCOM were all above 80 points, and the practicality was between 70 points and 90 points. Among them, the practicability of scene design and the reliability of text design were relatively high. The reason may be that the scene design is closer to real life, so it is more practical. The content of the text design is simpler, so it is more reliable. From the application effect of AI in Fig. 5B, the reliability scores of logo and text design were relatively high, about 90 points.

In comparison, the reliability scores of other design types were about 70 points. In practicality, poster design, text design and packaging design scored higher, above 80 points, while other design types scored around 70 points. This shows that most experts have recognized this essay through applying big data cluster analysis and AI identification, and it has certain reliability and practicability. This essay has achieved some results by applying big data clustering and AI recognition in VISCOM art design, but the application methods still need to be diversified. It is also necessary to deepen the application of enterprise technology to aid in developing artwork, which represents a rethinking of the Internet age and is a fresh style with contemporary qualities. Additionally, multimedia dialects and forms improve the number of artistic expressions. The exploration of VISCOM art makes it possible through AI technologies, enabling humans to use its features fairly and utilizing science as a springboard for investigating novel concepts, fresh aesthetics, and fresh creative methods. For the application of VISCOM art design, it is hoped that when big data and AI technology gradually become involved in the field of art, it can provide some new directions for the traditional visual art path. Based on the study’s results, the VISCOM art creation process makes excellent use of CNN and STING algorithms, which improve the accuracy of context- and audience-specific visual features. With this data-driven approach, designers can concentrate on originality and creativity, lessening the need for manual operations. If these algorithms work, they might be used for other forms of visual communication, including adaptive user interfaces and dynamic content production. The importance of integrating AI and big data technologies will grow as the area progresses.

Although this study offers vital insights into utilizing AI and big data in VISCOM visual design, it is important to accept certain limits to maintain a balanced perspective. An important constraint is the possibility of bias present in the datasets utilized. The data sources, although varied, may not comprehensively depict all potential design scenarios, especially those that involve extremely specialized or culturally distinctive information. This could result in biased outcomes, where the algorithms exhibit superior performance on some categories of data but may have difficulties with data that falls outside the dataset’s domain. Another constraint exists within the algorithms themselves. Although the CNN and STING algorithms demonstrated effectiveness in our trials, they possess intrinsic limits. The CNN algorithm is susceptible to overfitting when working with small datasets, while the STING algorithm may exhibit inefficiencies when applied to high-dimensional data. Furthermore, the effectiveness of these algorithms is greatly influenced by the caliber of the input data and the parameters used during training, which may not consistently apply to novel or unfamiliar material.

In addition, although seasoned experts carry out expert evaluations, they are nonetheless susceptible to personal bias and subjective judgment, which may impact the final rankings. Our findings in VISCOM may not apply to other areas of visual communication or diverse artistic styles due to the narrow focus on specific design tasks. By recognizing these constraints, we enhance our comprehension of the study’s results and propose recommendations for future research to tackle these difficulties.

Ethical considerations in AI-driven art and design

As artificial intelligence becomes more deeply incorporated into art and design, addressing ethical concerns related to authorship and originality is crucial. Authorship dilemmas arise when AI systems contribute substantially to creative works, questioning conventional ideas about who deserves recognition. Moreover, the authenticity of AI-generated art is called into doubt, as these systems frequently rely on preexisting works, potentially duplicating them. Furthermore, AI poses the potential danger of unintentional bias, as it may perpetuate cultural or stylistic biases already existing in the training data. It is crucial to address these ethical considerations to guarantee that AI’s involvement in art and design is responsible and shows proper respect for human creativity.