Categorization of digitized artworks by media with brain programming

Area of Science:

• Computer Science
• Artificial Intelligence
• Art History

Background:

• Art categorization by materials and techniques is a complex, open research challenge.
• Traditional object categorization methods may struggle with the visual diversity within art media categories.
• Machine learning offers potential solutions for identifying patterns in artistic techniques.

Purpose of the Study:

• To evaluate brain programming for art media categorization.
• To compare the efficacy of brain programming against deep learning in this domain.
• To identify the strengths and limitations of brain programming for art analysis.

Main Methods:

• Utilized brain programming, a machine learning technique, for art media classification.
• Developed and extracted feature-identifying functions to discern artistic patterns.
• Trained and validated models using the Kaggle database and tested on WikiArt.
• Benchmarked performance against established deep learning methodologies.

Main Results:

• Brain programming achieved over 90% accuracy in three out of five art classes.
• Performance was within 5% of deep learning in the remaining two classes.
• The brain programming approach utilized significantly simpler computational programs.

Conclusions:

• Brain programming demonstrates strong potential for art media categorization.
• This method offers a competitive and computationally efficient alternative to deep learning for art analysis.
• Further research can explore optimizing brain programming for nuanced art classification tasks.