Study of Chinese Shadow Mapping Classification with the Application of Deep Learning Algorithms

Area of Science:

• Computer Science
• Artificial Intelligence
• Digital Arts

Background:

• Shadow puppetry is a traditional Chinese art form, evolving into digital formats.
• Classifying digital shadow puppetry is challenging, requiring advanced vision-based solutions.
• The entertainment industry shows growing interest in digital shadow puppetry.

Purpose of the Study:

• To design an artificial intelligence-based classifier for digital shadow puppetry.
• To improve the accuracy and efficiency of identifying and classifying digital shadow puppet performances.
• To address the real-world vision-based problem of digital shadow puppet classification.

Main Methods:

• Utilized data augmentation to expand the training and testing datasets.
• Employed AlexNet, a deep neural network, for feature extraction from puppet images.
• Applied the Extreme Learning Classifier (ELC) for assigning class labels.
• Developed a modified Grey Wolf Optimized Classifier (mGWOC) for classification.

Main Results:

• The proposed mGWOC demonstrated superior performance compared to ResNet and DenseNet models.
• Achieved an average accuracy of 0.951.
• Outperformed the standard grey wolf optimization algorithm in precision, recall, F-score, and kappa statistics.

Conclusions:

• The mGWOC is an effective and accurate method for classifying digital shadow puppetry.
• This AI-driven approach enhances the identification process for digital shadow puppet performances.
• The research contributes a robust solution to a niche problem in the digital entertainment sector.