Abstract: - This paper will introduce the application of artificial intelligence (AI), machine learning, and deep

learning in art design and visual arts, and how these technologies can be used to create unique souvenirs. In the

field of art design, AI and machine learning can be used to automatically generate artwork and patterns,

providing more inspiration and creativity, and can also help artists better understand their audience and market.

The application of deep learning in the field of visual arts includes image recognition, image classification,

image generation, and so on. In the field of souvenir design, the use of AI and machine learning can help

designers better understand market needs and consumer trends to create unique souvenirs. Taken together, the

application of AI, machine learning, and deep learning technologies has great potential and creativity in the

The birth of the field of artificial intelligence (AI)

can be traced back to the Dartmouth Conference in

1956, when John McCarthy coined the term

systems or platforms are intelligent complexes

based on machine learning, [2], and deep learning

algorithms, [3]. The earliest machine-learning

algorithm model is the M-P model , [4]. The model

the connection weights by learning and can

correctly classify similar or different models.

However, the single-layer perceptron network

model cannot handle linear inseparability problems.

In 1986, the concept of the Back Propagation (BP)

learning begins by employing unsupervised learning

for the pre-training of each layer in the network.

One layer is trained at a time without supervision,

and the resulting training output from that layer is

subsequently utilized as the input for the next layer.

Finally, supervised learning is used to fine-tune the

pre-trained network, Convolutional neural network

(CNN), [9], is a feed-forward neural network

containing convolution operations that can identify

and classify images. A recurrent neural network

dimension while maintaining the characteristics of

the original data as much as possible.

hardware computing capabilities and the progress of

deep learning technology, the performance of many

computer vision and image processing tasks has

been greatly improved. Convolutional neural

networks have achieved success in vision tasks such

as image classification, [13], object detection, [14],

and semantic segmentation, [15]. The ImageNet

GoogLeNet [17], VGG [18], ResNet [19]. With the

emergence of these high-performance neural

networks, AI technology is constantly being pushed

forward.

in the field of art, it can stably and quickly help

designers complete time-consuming work in the

design process, saving a lot of time and greatly

improving work efficiency. Designers act more as

guides and perfecters in the creative process,

methods for traditional visual art creation forms but

also provide designers with more creativity. The

intervention of artificial intelligence in art creation

is not a replacement, but a joint development of man

and machine, and promotes art creation to move

organizes papers on artificial intelligence and art.

By searching and reading a large number of papers,

to determine the research purpose and significance

of this paper. And discover the existing problems,

and grasp the focus of this paper. Focus on the

research and analysis of artificial intelligence in

visual design. Explore the help and significance of

artificial intelligence to visual art creation in the

digital age.

directions, which effectively avoids the subjective

tendency of researchers in different research

backgrounds, [25]. The method of case analysis is a

systematic approach that entails conducting

thorough and detailed research on representative

entities to achieve a comprehensive understanding,

[26]. This paper collects and screens representative

cases of the integration of artificial intelligence and

art design for further in-depth analysis. Based on the

summary of actual cases, it explains the new form of

artificial intelligence and visual art.

2.3 The Experimental Analysis Method

The experimental analysis method refers to the

design of simulation experiments to reproduce the

various factors and development processes of

things, to find various useful data about such things

in the real world, [27]. The realization of intelligent

with the help of Deep Dream, [29]. As shown in

Figure 2, the creator chose the Google map plan of

the British intelligence and security agency, input it

into the Deep Dream Generator, generated multiple

morphed versions, and finally fused the multiple

versions into the final work.

shocking visual effects. It eventually sold for

$8,000, a positive affirmation of the collaborative

creative model of humans and artificial intelligence.

find similar works.

Fig. 4: Bear stick figure recognition based on the

AutoDraw tool

4.1 Art Creation Process based on AI

Traditional art design does not rely on computers or

artificial intelligence, and highlights human

maximum value or the average value. This operation

is to downsample the feature map to reduce the size

of the feature map. Each neuron in the fully

connected layer of CNN is connected to all neurons

in the previous layer, enabling the network to learn

features and output each category. Each neuron in

the fully connected layer corresponds to a specific

class label, thereby achieving accurate

classification.

convolutional neural network architecture. As

shown in Figure 6 and Figure 7, the VGG19

structure consists of a series of convolutional layers,

pooling layers, and fully connected layers. VGG19

has the characteristics of a small convolution kernel

and a small pooling kernel. The size of the

convolution kernel is 3×3 and the size of the pooling

layer is 2×2.

Fig. 6: The overall frame diagram of VGG19

loss and style loss. Content loss is used to preserve

the structure and details of the content image, while

style loss is used to preserve the texture and color

features of the style image. The overall loss function

is obtained by a weighted summation of the two

losses.

overall loss function for generating images. During

optimization, the pixel values of the generative

possesses the content structure and detail of the

content image and the texture and color

characteristics of the style image.

4.6 Result

Applying the above method, we transferred the style

of Figure 8 and Figure 9, and the result is shown in

Figure 10. Figure 10 has the original picture as the

background with the artistic style of Van Gogh's

Starry Night applied to it. The resulting picture is

more psychedelic and dizzying, giving it a more

artistic feel. By blending the content of Figure 10

with the style of Starry Night, the resulting image

combines the structure and details of the original

image with the texture and color characteristics of

Fig. 8: Stylistic drawing (Van Gogh's work: Starry

Night)

in depth how to use AI technology for artistic

creation. First, by introducing relevant cases and

conducting in-depth analysis, we aim to gain a more

comprehensive understanding of the role of artificial

intelligence in the art field. Next, this article shows

a specific process of using AI technology for artistic

creation. Specifically, we used the VGG19 neural

network and Adam optimizer to perform style

conversion on the original artistic pictures through

transfer learning, using pictures of a specific style as

technology for artistic creation. Developments in

this field are expected to provide more creative

inspiration and tools for art creators.

References:

[1] Rajaraman V. JohnMcCarthy—Father of

[2] Jordan MI, Mitchell TM. Machine learning:

[16] Alom MZ, Taha TM, Yakopcic C, Westberg

[17] Al-Qizwini M, Barjasteh I, Al-Qassab H,

[19] Targ S, Almeida D, Lyman K. Resnet in

[20] Han Z. Combined Development of Artificial

[21] Ye C. Evolution and Application of Artificial

[23] Lin Y-C, Chen Y-T. Artificial intelligent

[26] Seawright J, Gerring J. Case selection