Transformation of Plants into Polka Dot Arts: Kusama Yayoi as an Inspiration for Deep Learning.

Museums and galleries are places that stimulate the imagination and creativity of people. However, against the backdrop of the COVID-19 pandemic, many museums have been forced to close. As a result, more visitors have had to start visiting online exhibitions. Under this trend, the quality of the online exhibition experience is facing unprecedented challenges, and visitors need to be provided with a high-quality online experience. With the rapid development of emerging technologies, museums and galleries have started to use technology to create "digitalization" experiences, such as 3D virtual exhibitions and VR interactive experiences. The current online exhibition format focuses more on improving the visitor's experience of viewing and browsing works and related information and less on how visitors create their artworks. In order to remedy the experience of visitors actively creating artworks, this study uses Yayoi Kusama's works as a case study to help form an utterly interactive process between visitors and artworks. We propose a Polka Dot Arts Transfer Network (PDAT-net), mixing realisticlooking and Polka Dot Art-style reference images. The output image looks like an actual natural image "drawn" in the polka dot art style. The PDAT-net is trained to compose an image in the style of Polka Dot Arts, based on neural style transfer. Convolutional networks implemented the training. Plant images performed the final test. The results showed that the generated polka dot images have similar styles to the style images, proving that our PDAT-network has good transfer performance of Polka Dot Arts.