We want to make skin cancer diagnosis available to everyone in the world, especially people in third world countries who don't have access to doctors and dermatologists.
Figure 1: This statistical visualization of diagnosis age for malignant tumours stood out to us, and was one of the reasons we decided it was a problem that needed to be diagnosed early on.
What it does and how we built it
e-Diagnose was trained on a dataset of labelled malignant and benign skin cancer tumours. It was retrained with Google's TensorFlow Inception model.
Challenges we ran into
Understanding the mathematical and theoretical aspects of how neural networks work. We had troubles grasping the need to use a docker container, but quickly we realized it was very effective.
Accomplishments that we're proud of
We're proud of achieving a similar classification accuracy to real-world skin cancer diagnosis.
What's next for e-Diagnose
We plan to implement our algorithm and training model on a front-end system, by creating a cross-platform mobile application that will let anyone take a picture of a suspected tumour.
Installation and Project Setup
1. Install Docker Toolbox for your machine.
2. Clone the repository.
git clone https://github.com/paarthmadan/skin-cancer-cnn
3. Use the trained model, or retrain the model by running retrain.py
4. Run main.py