Skin diseases are a common problems and affect about 20-25% of global population annually [1]. However, due to dermatologist shortage, patients suffer long waiting time for scheduling an appointment (averaging 36 days in the US) [2]. This causes a significant amount of stress in patients and delay in diagnosis and treatment of the disease. AI-powered tools can assist patients to perform preliminary diagnosis and apply home remedies till they visit the dermatologist.
DermAssist contains a powerful vision model to classify skin disease type given the image of the affected area, and a Retrieval-Augmented Generation (RAG) system enriched with reliable dermatological knowledge to conduct a conversation and suggest remedies.
Clone the repository:
https://github.com/chinmaysharmacs10/DermAssist.git
cd DermAssist
Create virtual environment:
Conda:
conda create --name dermassist
Pip:
python -m venv dermassist
Install necessary packages:
pip install requirements.txt
Download ollama client from: https://ollama.com
Pull Llama3-8b model:
ollama run llama3
In rag_system.py, enter your LANGCHAIN_API_KEY and LANGCHAIN_PROJECT to enable tracing via LangSmith.
To start StreamLit server:
streamlit run dermassist_streamlit.py
- Xuhui Li, Xinyu Zhao, Haoran Ma, and Bin Xie. Image analysis and diagnosis of skin diseases-a review. Current Medical Imaging, 19(3):199–242, 2023.
- Elizabeth Uhlenhake, Robert Brodell, and Eliot Mostow. The dermatology work force: a focus on urban versus rural wait times. Journal of the American Academy of Dermatology, 61(1):17–22, 2009.