VisionGuard: Your Early Detection Shield for Eye Diseases

Inspiration

Hearing stories of individuals who battled eye cancer and the challenges they faced in late-stage diagnoses was a driving force. These experiences highlighted the need for a proactive, accessible, and non-invasive solution.

What it does

It uses images taken from phone and using an AI model it detects the early signs of any disease.

How we built it

we use a machine learning model (CNN) to identify and extract disease-related information from Ocular images taken by a phone. We then use the extracted information to identify four conditions: Normal, Diabetic, Glaucoma, Cataract.

Challenges we ran into

Finding a clean and useful data, training the model to achieve a good accuracy, deploying the model to be accessible to the user

Accomplishments that we're proud of

High accuracy detection (94% test accuracy), using limited technology (phone and a magnifier and mydriatic drugs to perform accurate early eye disease detection.

What we learned

team work, time management, using AI to facilitate our work, working under pressure

What's next for VisionGuard: Your Early Detection Shield for Eye Cancer

We will start making it accessible to people in low income regions of the world, we then use the collected information to garner a high quality dataset of images that will help us improve the existing model. We also aim to integrate other modalities into our model, e.g., using skin color and nail colors, and lips to identify different set of disease or to only provide more robust early evaluation.

Built With

• python
• taipy
• tensorflow
• tensorflowlite