Inspiration
Skin cancers like melanoma start off as subtle changes to your skin complexion, or even lesions. Many forms of skin cancer is treatable if it's found early. However, it is very inconvenient & expensive to get diagnosed for Skin Cancer.
What it does
App that offers a more complete and reliable experience for users to detect potentially malignant skin cancer using CNN to determine if the cancer is malignant or benign. Users can simply use the application and upload pictures of the area of skin that they suspect is cancerous. The application will determine if it is potentially malignant or benign.
How we built it
With datasets obtained from Kaggle, we used Tensorflow for the machine learning, and built a Convolution Neural Network. The model performance at 30 epochs was around 80%, with a training loss of around 30%. After which, we made use of Flask to build a web application that could serve up the model and do some predictions based on images uploaded. The images uploaded by the user is also saved into a specific folder.
Challenges we ran into
Time challenge with the training of the model. Challenge to find a suitable dataset. Packaging the whole application as a web app.
Accomplishments that we're proud of
Being able to build a very accurate model that is able to discern between malignant and benign cells. And delivering it in a simple web application. Building an NLP based question answering model.
What we learned
We learnt how important it is to adjust our hyperparameters, we learnt how to deploy a machine learning model to be used with a web application. Importance of coming up with solutions based on existing problems.
What's next for CAmcRE by Masrine Learning
Offer a chatbot that gives valuable information on skin cancer, quick health tips to lower risk of skin cancer. Make use other applications such as Google Maps to offer directions to the nearest dermatologist. Offer options to book appointments with dermatologists. Deploy the application as a mobile app that users can use on the go.
Log in or sign up for Devpost to join the conversation.