Exploring Color Models for Forensic Hair Comparisons: A Machine Learning Perspective

This study investigates the creation of a sophisticated system for classifying hair types through the use of a hybrid technique that combines Random Forest algorithms and Convolutional Neural Networks (CNNs). Ten distinct hair texture types-Straight, Wavy, Straight Curls, Loose Curls, Spiral Curls, Kinky, Coiled, Corkscrew Curls, Afro-textured, and Ringlets-were used in our training and evaluation dataset. The model's recall, precision, and F'l-Score ranged from 93.35% to 99%, indicating its outstanding accuracy. The model exhibited a noteworthy 84<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> overall accuracy rate for the majority of hair texture classes, underscoring its efficacy in precisely classifying diverse hair kinds. To increase the “Ringlets” hair type's classification accuracy, more refinement is necessary. In the classification of hair types, our hybrid approach-which incorporates Random Forest algorithms with Convolutional Neural Networks-proved to be extremely dependable and effective. The study advances automated systems for accurate and consistent classification of hair types, which could find use in several sectors such as dermatology, fashion, and cosmetics. The results highlight how our hybrid approach effectively addresses the drawbacks of manual classification approaches by improving the effectiveness and precision of computerized hair type categorization systems.