Skinsight

We want to make dermatology more accessible to everyone. Users can upload images of skin marks to be classified as benign or malignant, as well as by subtype. We used the st-cropper library to allow the user to crop their image before predictions. We used a Dual Head ResNet-18 model for quick but in-depth results. This also allowed for both binary and multi-class classification.

Built With

• isic
• python
• pytorch