Skin Lesions Classification Using Deep Ensemble Learning Model of VGG-16, ResNet-50 and Inception-V3

• Author(s): Shubham Chaurasia ; Shiwangi Choudhary
• Paper ID: 1708324
• Page: 284-295
• Published Date: 08-05-2025
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 8 Issue 11 May-2025

Abstract

This research introduces a sophisticated methodology for the classification of multi-class skin lesions by utilizing an ensemble model that integrates the Inception-V3, ResNet-50, and VGG16 architectures. The primary objective of the classification task is to categorize skin lesions into specific classes, such as Melanoma (MEL), basal cell carcinoma (BCC), and Nevus (NEV), employing the ISIC dataset, which is a comprehensive repository of dermoscopic images. To address the issue of dataset imbalance, an oversampling strategy is implemented, as certain lesion types are underrepresented. This approach ensures that the training of model is done on a more balanced dataset, thereby enhancing its ability to classify all lesion types accurately and fairly. The ensemble technique capitalizes on the unique strengths of ResNet-50, Inception-V3, and VGG16, leading to improved overall classification performance. ResNet-50 is selected for its proficiency in deep feature extraction, which is essential for capturing intricate details in lesion patterns. Inception-V3 is chosen for its capability to process lesions at multiple scales, effectively analyzing variations in resolution and size. VGG16 is incorporated due to its straightforward yet highly efficient architecture for image classification tasks. The ensemble model, augmented with data enhancement techniques, significantly surpasses the performance of individual models in skin lesion classification, achieving superior results in precision, accuracy, recall, and F1-score for both the original and balanced ISIC datasets. This approach provides a robust framework for skin lesion classification, thereby enhancing the reliability and accuracy of diagnostic process in the field of dermatology.

Keywords

Skin Lesion Classification, Ensemble Model, ResNet-50, VGG-16, Inception-V3

Citations

IRE Journals:
Shubham Chaurasia , Shiwangi Choudhary "Skin Lesions Classification Using Deep Ensemble Learning Model of VGG-16, ResNet-50 and Inception-V3" Iconic Research And Engineering Journals Volume 8 Issue 11 2025 Page 284-295

IEEE:
Shubham Chaurasia , Shiwangi Choudhary "Skin Lesions Classification Using Deep Ensemble Learning Model of VGG-16, ResNet-50 and Inception-V3" Iconic Research And Engineering Journals, 8(11)