A DEEP CONVOLUTIONAL NEURAL NETWORK APPROACH FOR MULTI-CLASS SKIN DISEASE RECOGNITION

Abstract

Skin problems are super common globally and can cause big issues if not caught and handled fast. Because we need better ways to find these problems, machine learning is becoming a good way to give doctors a hand in spotting them early. In this study, we made a machine learning setup that uses a souped-up ResNet50 design to sort different kinds of skin diseases. We got the data ready and split it into training, validation, and testing groups with a 90-5-5 split. We also beefed up the data a lot to make it work better across the board. The model has special layers to keep it from being too specific and to keep the training steady, and the way it learns changes to get the best results. Our model scored about 90% on the tests, which is good for spotting seven kinds of skin issues. The results show that the ResNet50 method is good at spotting tricky visual things in medical images, which can help doctors give quicker and more spot-on diagnoses. This shows that AI could be a game-changer for looking at medical images and making skin care more available, trustworthy, and based on data.