Unmasking Comic Narratives: CNN-RF Approach to Genre Classification

Classifying comic genres is challenging in computer vision and image analysis, with broad implications for image retrieval, comic book cataloguing, and content recommendation. In this study, the authors provide a novel approach designed for the automatic genre categorization of comic pictures. Our method aims to accomplish accurate and complex genre identification using the powerful synergy between Convolutional Neural Networks (CNNs) and a Random Forest (RF) classifier. Carefully calculated for every genre category, important metrics are used to assess the model's performance: precision, recall, F1-score, support, and overall accuracy. The testing results show impressive skill, exhibiting high categorization accuracy in all genres. With a 98% accuracy rate overall, this is rather outstanding. More specifically, the model performs very well in accuracy, recall, and F1 scores for the Humor, Battle, and Love Fantasy genres, repeatedly exceeding 90% in these critical measures. Interestingly, the genres of Science Fiction, Romance, Horror, Action, and Science Fiction all do well in terms of categorization, with F1 scores that often above 90%. The CNN- RF model has shown a strong and all-encompassing performance that highlights its potential to transform the automation of comic genre categorization tasks.