In my quest to find a problem to solve in my closing year at college, I chanced upon an online write up by a reputable newspaper which stated that there were only about 25 certified dermatologists in Ghana as at 2017, which was to serve a population of about 25 million at the time. Really!! I very much intrigued by this news. I imagined how the nation would cope if there was an outbreak of a skin disease, coupled with the lack of medical equipment. As a result I teamed up with a colleague of tackle the problem.
The goal of the project was to apply artificial intelligence and computer vision using deep learning techniques to aid physicians detect the deadliest type of skin cancer, melanoma, very early. We developed a deep learning model to accurately classify skin lesions as either benign or malignant using Transfer Learning.
Training and validation statistics
Deployed via Streamlit
- Python
- TensorFlow
- Pandas
- Matplotlib
- OpenCV
- Streamlit
My deepest gratitude goes to Cherif for this great collaboration.