Skin Disease Classification

Real-time-Detector

Inspiration

I was inspired to create this project because I had previously created a skin-cancer detector, and I wanted to expand and create a more useful application. Thus, I created a web application that the user can use to classify what a skin mark on their hand might be. I knew a web application would be more useful to regular people and that such an application will save people time from dermatologist appointments if they have harmless skin conditions.

What it does

This application allows the user to capture an image of a skin mark on their body, and it predicts what the mark may be into 1 of 10 skin-disease categories. The web app also consists of hyperlinks so that the user can look into what the predicted skin mark is and take the right course of action.

How we built it

To build the application, I primarily used Python. I created the front-end using HTML and CSS, and I also used Flask to create a backend. To classify the image that the user captured and predict what it is, I used TensorFlow and made my own Convolutional Neural Network with approximately 75% accuracy. I also used OpenCV to display a webcam on the web application and allow the user to capture an image in real time.

Challenges we ran into

Some challenges I ran into included passing the captured image into the neural network and displaying the output on the web application. I initially struggled to display the prediction because of the way my code was structured, but I managed to fix the issue.

Accomplishments that we're proud of

I'm most proud of the fact that I created my own convolutional neural network and then connected it to a web application, as this was the first time that I created such a complex application by myself.

What we learned

I learned about image pre-processing in TensorFlow as I was creating this project because this is the first time that I created a Convolutional Neural Network. I also learned how I could display the prediction of a machine learning model on a web application.

What's next for Skin Disease Detector

To improve and expand on this project, I would improve the user interface of the web application and create a more accurate machine learning model (planning to reach around 90% accuracy). I would also like to add more skin conditions that the model could classify.