Wound care

Inspiration

I was inspired by a project (AI Agent on cancer analysis) presented in a hackathon prepared in Addis Ababa.

What it does

It classifies wound image and process medical image after uploading and then answers question based on the context.

How we built it

I used vit transformer based model from hugging face that was fine-tunned for wound image classification and gemini api for question answering. I built the frontend using react+vite and the backend using django

Challenges we ran into

At first I fine-tunned a resnet based model the accuracy was not much good. So I change it to vit transformer model. Online vector database store was high.

Accomplishments that we're proud of

Integrating these model to a chatbot. And being able to ask question from a document and image. working in a real world case

What we learned

I learned how I can use vision transformer to enhance the skill on image recognition and classification.

What's next for Wound care

enhance the web by improving the model accuracy. and fine-tunning LLM to answer question with out uploading files .

Built With

• django
• gemini
• hugging-face
• vite+react