Cargando…

Application of Artificial Intelligence within Virtual Reality for Production of Digital Media Art

As technology changes, virtual reality generates realistic images through computer graphics and provides users with an immersive experience through various interactive means. In the context of digitalization, the application of VR for digital media art creation becomes a normalized method. Today...

Descripción completa

Detalles Bibliográficos
Autor principal: Wu, Yunxuan
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Hindawi 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9385317/
https://www.ncbi.nlm.nih.gov/pubmed/35990155
http://dx.doi.org/10.1155/2022/3781750
Descripción
Sumario:As technology changes, virtual reality generates realistic images through computer graphics and provides users with an immersive experience through various interactive means. In the context of digitalization, the application of VR for digital media art creation becomes a normalized method. Today's digital media art creation is closely related to vigorous technological innovation behind it, so the influence of modern technology is inevitable. Virtual reality and artificial intelligence have gradually become the main technical means in line with the development aim for digital media art creation. This work proposes an art object detection method AODNET in virtual reality digital media art creation with AI. Aiming at the particularity of object detection in this direction, an art object detection strategy based on residual network and clustering idea is proposed. First of all, it uses ResNet50 as backbone, which deepens network depth and improves the model feature extraction ability. Second, it uses the K-means++ algorithm to perform clustering statistics on the size of the real annotated boxes in the dataset to obtain appropriate hyperparameters for preset candidate boxes, which enhances the tolerance of the algorithm to the target size. Third, it replaces the ROI pooling algorithm with ROI align to eliminate the error caused by the quantization operation on the characteristics of the candidate region. Fourth, to reduce the missed detection rate of overlapping targets, soft-NMS algorithm is used instead of the NMS algorithm to post-process the candidate boxes. Finally, this work conducts extensive experiments to verify the superiority of AODNET for object detection in virtual reality digital media art creation.