Authors:
A. Muthukumaravel, S. Silvia Priscila, B. M. Praveen
Addresses:
Institute of Engineering and Technology, Srinivas University, Dakshina Kannada, Karnataka, India. Faculty of Arts and Science, Bharath Institute of Higher Education and Research, Chennai, Tamil Nadu, India. Department of Computer Science, Bharath Institute of Higher Education and Research, Chennai, Tamil Nadu, India. Institute of Engineering and Technology, Srinivas University, Dakshina Kannada, Karnataka, India.
Abstract:
In view of the increasing dermatological disease as well as the inevitable shortage of qualified experts, the automated skin syndrome classification based on DL (Deep Learning) has gained much significance recently. The research considers a CNN (Convolutional Neural Network) model for classifying multi-class skin diseases, compares its performance with existing DL models such as VGG16 and ResNet50, and Presents Results on InceptionV3. The International Skin Imaging Collaboration (ISIC) dataset is used to evaluate the proposed approach, which consists of dermoscopic images from multiple classes. In the preprocessing stage, the image is resized, and the CLAHE (Contrast Limited Adaptive Histogram Equalisation) technique is applied to enhance lesion visibility while reducing illumination variation. The proposed HFF-CNN uses a hybrid feature fusion technique that concatenates the features learned from the custom-designed CNN with the high-level TL (Transfer Learning) features extracted from the pre-trained ResNet50 model. Hybrid features capture local texture details via a custom CNN and global semantic representations via a pre-trained CNN. The classification process takes place in the fully connected layers, using a concatenated feature vector and drop-out regularisation. The suggested HFF-CNN model is tested using accuracy, precision, recall, F1 score, and ROC curve area. The hybrid model beats all DL methods in classification accuracy and robustness, according to trials. The HFF-CNN architecture appears useful for early, accurate clinical diagnosis of skin disease.
Keywords: Skin Disease; Framework and Classification; Feature Fusion; Image Preprocessing; Deep Learning; Transfer Learning; Hybrid Model; Clinical Setting; Precision Value.
Received on: 23/11/2024, Revised on: 12/01/2025, Accepted on: 11/04/2025, Published on: 03/01/2026
DOI: 10.64091/ATICL.2026.000252
AVE Trends in Intelligent Computer Letters, 2026 Vol. 2 No. 1 , Pages: 13–25