Deep Learning Approaches to Babylonian Numeral Recognition: A CNN-SVM Hybrid Model

This paper presents a novel deep learning framework for classifying Babylonian numerals by integrating Convolutional Neural Networks (CNNs) with a hybrid CNN-SVM model. The core contributions include the development of a robust feature extraction pipeline combining zoning, component analysis, and angular analysis of cuneiform symbols, optimized with deep learning techniques. The proposed system extends beyond numeral classification by incorporating object recognition and counting mechanisms to accurately convert multi-digit Babylonian numbers into their decimal equivalents in the base- 60 system. This approach achieves remarkable accuracy of $98.33 \%$, significantly advancing automated recognition and interpretation of ancient mathematical symbols. By bridging machine learning with historical data analysis, the research contributes to the preservation and scholarly study of Babylonian mathematical heritage while demonstrating the efficacy of hybrid models for complex, historical numeral recognition tasks.