A Reliable and Effective Approach for Computerized Skin Disease Classification Using MobileNetV3 and LSTM

Deep learning models are highly effective at identifying features that facilitate the precise interpretation of complex patterns. This study introduces a deep learning-based approach using MobileNetV3 and long short-term memory (LSTM) for the automated classification of skin diseases. The MobileNetV3 model, which is compatible with mobile computing devices, has demonstrated both efficiency and reliability. The proposed model excels in maintaining stateful data for accurate weather forecasting. A grey-level co-occurrence ma-trix was employed to evaluate the progression of abnormal growth. The model's perfor-mance was compared with other advanced models, including convolutional neural net-works (CNN), very deep convolutional networks for large-scale image recognition devel-oped by the Visual Geometry Group (VGG), and fine-tuned neural networks (FTNN). For this study, we utilized the DERMNET dataset. Its minimal computational demand is at-tributed to its robustness in detecting the affected area, achieving this significantly faster than the standard MobileNet model with approximately two fewer calculations. The findings indicate that the proposed method can aid general practitioners in accurately diagnosing skin disorders, thereby reducing the risk of subsequent complications and morbidity in patients.