A Deep Learning Framework for Enhanced Skin Cancer Diagnosis using Hybrid VGG16-U-Net Architecture (KEY IJP************659)

• N.swathi,Mrs. A. Andrine Dinola M.e.,,M. Yamunashri,P. Yazhini

Abstract

Precise identification and timely recognition of malignant skin disorders are vigorous for enhancing treatment upshots and increasing persistence rates. The proposed system utilizes a symmetric encoder-decoder structure to efficiently segment skin lesions from dermoscopic images, enabling precise identification of malignant regions. In order to preserve fine-grained information for precise lesion localization, the encoder downsamples key characteristics while the decoder upsamples the segmented output. Additionally, an ESP32-CAM module is incorporated for real-time image capture and transmission, enhancing accessibility for telemedicine applications. The suggested solution overcomes the drawbacks of conventional diagnostic techniques, which are frequently laborious and subject to human mistake. A comparison with current approaches demonstrates our model's advantage providing computing efficiency and segmentation accuracy. The usefulness of the system is estimated using key metrics that measure its precision in identifying affected areas and segmenting images accurately. These indicators demonstrate the system's ability to enhance the timely detection of malignant skin conditions. Combining cutting-edge deep learning methods with real-time imaging capabilities offers a scalable and effective way to diagnose dermatological conditions. This research has significant implications for telemedicine and clinical practices, offering a cost-effective and automated approach to skin cancer detection.Keywords skin cancer, Machine learning, CNN, U- Net, detection, ESP32-CAM.