The purpose of this project is to mitigate doctor's task to segment various malignant tumors on human's skin.
We use neural network, namely UNet, for image segmentation. Furthermore, sequential neural network was trained on labelled dataset of 9 types of skin tumors for 30 epochs and achieved almost 80% of accuracy on test dataset.
Here is the input image:
Here is the output image:
