What's That Rash?

• 

  Skin Condition Prediction Page

• 

  Skincare Routine Recommendation Page

Inspiration

We realized that identifying skin conditions can be confusing due to the overwhelming variety, from minor allergic rashes to more serious dermatological issues like psoriasis or eczema. Many people turn to unreliable internet searches or delay treatment due to uncertainty. We saw an opportunity to build a streamlined, AI-powered tool to simplify this process and make accurate skin assessments more accessible.

What it does

What's That Rash? is a web app that helps users identify skin conditions with just a photo. Users upload an image of their skin issue (such as a rash, bump, or bruise) and add descriptive tags. Our AI dermatologist analyzes the input and provides a likely diagnosis. Users can then generate a professional PDF report and email it directly to a doctor.

Additionally, the app offers personalized skincare routines based on the user’s skin type and concerns, with a link to product recommendations, blending medical insight with daily care.

How we built it

We used the Streamlit framework for rapid prototyping and OpenAI’s GPT-4 Vision API for image interpretation. To improve medical accuracy, we reinforced the system using a curated Skin Disease Classification dataset and fine-tuned it for dermatological relevance.

Challenges we ran into

Training and integrating the dataset was challenging. The dataset lacked diversity and volume, which affected diagnostic performance. We had to strike a balance between accuracy and generalizability while keeping the app accessible.

Accomplishments we're proud of

• Successfully integrated Streamlit and GPT-4 Vision into a fully functioning app, despite being new to both tools.
• Reinforced our model using additional datasets to improve prediction reliability.
• Designed a clean and intuitive interface within a limited timeframe.
• We also validated our app through user testing to ensure that the output was both useful and understandable.

What we learned

On the development side, we learned how to work with the Streamlit ecosystem, implement image-based AI using Vision APIs, and manage dataset integration. From a collaboration standpoint, we gained insight into how dataset quality impacts AI performance. We also developed effective workflows while coordinating across different time zones, with members based in both Singapore and the United States.

What's next for What's That Rash?

We plan to refine our model using higher-quality and more diverse dermatology datasets. We aim to improve user experience by enabling automatic PDF downloads and email attachments. We also want to implement multilingual support to improve accessibility across different regions and demographics.

Built With

• gpt
• openai
• python
• streamlit