Hairlossmultinet: A Multi-Scale Feature Fusion Method Using Deep Learning Approach

    January 2025 in “ Communications in computer and information science ”
    Nizamul Haque Sohan, Md Mahbubur Rahman, Ahmed Shafkat, Bijon Mallik, Ahsan Mahbub, Nazmul Hassan
    TLDR HairLossMultinet accurately classifies hair damage with 98% accuracy but needs a more diverse dataset for broader use.
    The paper introduces HairLossMultinet, a hybrid deep learning framework that combines ResNet50 and VGG19 to enhance hair damage classification accuracy. By employing a multi-scale feature fusion scheme, the model integrates high- and low-level visual patterns, achieving a 98% accuracy rate on a dataset of 5,304 images labeled as 'Damaged' or 'Normal'. The study highlights the model's superior performance compared to eight state-of-the-art convolutional neural networks and uses Grad-CAM to validate predictions. However, the limited diversity of the dataset poses a challenge to clinical generalizability. This research lays the groundwork for future advancements in explainable AI for dermatological diagnostics.
    Discuss this study in the Community →

    Research cited in this study

    4 / 4 results