Transfer learning approaches for EfficientNetV2 B0 and ImageNet skin cancer classification in a convolutional neural network

Early skin cancer detection by visual inspection of skin lesion images may be challenging. In recent years, there have been impressive findings from research into the use of deep learning algorithms to assist in identifying skin cancer. Modern techniques are more sensitive, specific, and accurate than dermatologists. However, dermoscopy image analysis using deep learning models still faces some challenges. Image segmentation, noise filtering, and inconsistent capture settings are all examples of such tasks. Melanoma, the deadliest form of skin cancer, is on the rise in occurrence. Early detection is crucial for preventing the spread of skin cancer. An automated technology that can detect skin cancer independently of human doctors is introduced in this study. Everyone offers a transfer learning-based approach to melanoma lesion detection. The images taken with the proposed technology are first preprocessed to remove noise and lighting influences. An important part of deep learning is image augmentation. Here are a few ways to make your images seem better. Image augmentation, which involves enlarging the training images, may improve the learning phase’s effectiveness. An innovative deep transfer learning technique for melanoma classification is proposed in this article using EfficientNet V2 B0 and ImageNet. To identify malignant skin lesions in a sample, two deep convolutional neural networks may be used: MobileNetV2 and ImageNet. We test the proposed deep learning model on the ISIC 2020 dataset. The suspicious spot in the middle of the image becomes the focus point at each stage. The combined results from these parts are put into a fully connected neural network. Results from experiments demonstrate that the proposed technique outperforms state-of-the-art deep learning algorithms in terms of accuracy while using fewer computational resources.