Hi all,
I am working on a skin disease classification project. I have image data for 8 different skin diseases, and I performed transfer learning using the ResNet50 model, but I am only achieving 70% validation accuracy. I also used data augmentation to increase the accuracy, but I have not had any success. Could anyone please help me and provide some ideas on how to improve the accuracy?
Perhaps 70% accuracy is the best you can do with this set of data and that model.
Try using label studio for annotation, if you have not done it properly yet, and try with hybrid models to capture features better first and then use models like lstm to classify…