Enhancing Skin Lesion Classification Accuracy: A Deep Learning Approach Using RexNet 150 for Improved Dermatological Diagnosis

The diagnosis of skin lesions is still a major global health concern, requiring the use of sophisticated procedures for early and precise identification. In this work, we suggest a deep learning strategy that makes use of the complex RexNet 150 architecture to improve the segmentation accuracy of skin lesions. Utilizing a broad dataset of more than 60,000 photos representing a range of skin lesions, such as Melanoma, BCC and Keratoses, our approach attains exceptional testing accuracy of 97.91% and training accuracy of 99.21%. RexNet 150 outperforms conventional CNN models, exhibiting markedly improved accuracy and reduced loss, proving its effectiveness in correctly classifying skin lesions. Our goal is to enhance skin cancer treatment and early diagnosis by utilizing RexNet 150's capabilities, which will aid in the continuous fight against this common and potentially lethal illness.