SKINBOX - DERMATOLOGY CONSULTING CHATBOT

Developed by: STACKLIFO TEAM

1. PROJECT INTRODUCTION

1.1 Overview and Context

In the community today, skin issues such as acne, allergies, and dermatitis are increasingly common, directly impacting the health and mental well-being of many people. However, not everyone has the conditions to visit specialized dermatology facilities directly. Many individuals choose to self-study information or post images of their affected skin online for opinions. The information received is often inaccurate, which can even lead to serious consequences due to misdiagnosis or self-treatment.

1.2 Requirements and Goals

Given this reality, the project demands a technological solution to help people access accurate and safe medical information.

The specific requirement is to develop an Artificial Intelligence (AI) application capable of:

1. Image recognition of the damaged skin area.
2. Symptom analysis described by the user.
3. Providing reliable preliminary advice.

The project aims to develop an intelligent support tool, enabling users to initially assess their skin condition and subsequently determine the appropriate course of action, such as self-care or consulting a specialist when necessary.

2. TECHNOLOGY STACK

The chatbot system leverages several advanced technologies:

Technology	Role in the System
FLASK	Serves as the deployment platform and intermediate web server between users and the core AI system.
RAG (Retrieval-Augmented Generation)	Used with LangChain for Natural Language Processing (NLP), enabling the chatbot to understand context and generate accurate, natural responses.
CNN (Convolutional Neural Network)	Analyzes skin images submitted by users to identify lesions and suggest preliminary consultation.
MySQL	Manages the storage of user data and chat history, ensuring system stability and personalized user experience.
Qdrant	Functions as the Vector Database (VectorDB) for storing and rapidly retrieving semantic information.
Selenium	Used as a tool to collect the knowledge dataset from reputable medical sources.

3. SYSTEM ARCHITECTURE AND PERFORMANCE

3.1 Architecture

The system is designed with two primary processing flows:

1. Image Processing Flow (CNN): Analyzes user-uploaded skin images to identify pathological features.
2. Text Processing Flow (RAG): Analyzes user-described symptoms and retrieves information from the knowledge base.

The diagram below illustrates the RAG architecture used for the text processing flow:

(Note: The diagram uses "Pinecone" as an example VectorDB. This project uses "Qdrant" in the equivalent role.)

3.2 Key Results and Performance

• CNN Model Training Accuracy: The accuracy reached 71.5%.
• Dataset: The CNN model was trained on a dataset containing 272,000 images covering 10 disease labels.

3.3 Product Interfaces

The developed graphical user interfaces (GUIs) include:

1. Login and Registration Interface
2. Homepage Interface
3. Chatbot Interface
4. Contribution Page Interface

(Recommended: Insert screenshots of your application here, e.g., [screenshot_chat.png])

4. GETTING STARTED

This guide will get you a copy of the project up and running on your local machine for development and testing purposes.

4.1 Prerequisites

Ensure you have the following software installed:

• Python [e.g., 3.10+]
• MySQL Server [e.g., 8.0+]
• Qdrant (running as a service or Docker container)
• Git

4.2 Installation

1. Clone the repository:

   git clone [your-repository-url.git]
   cd [your-project-directory]

2. Create and activate a virtual environment (Recommended):

   # For macOS/Linux
   python3 -m venv venv
   source venv/bin/activate
   
   # For Windows
   python -m venv venv
   .\venv\Scripts\activate

3. Install the required packages:

   pip install -r requirements.txt

4.3 Configuration

1. Database Setup:

   • Access your MySQL server and create a new database for the project.
   • Example: CREATE DATABASE skinbox_db;

2. Environment Variables:

   • Create a .env file in the root directory of the project.
   • Copy the contents from .env.example (if it exists) or add the necessary variables.

   Example .env file:

   # Flask Configuration
   FLASK_APP=app.py
   FLASK_ENV=development
   SECRET_KEY=[your-very-long-random-secret-key]
   
   # Database Configuration
   DB_HOST=localhost
   DB_USER=root
   DB_PASSWORD=[your-mysql-password]
   DB_NAME=skinbox_db
   
   # Qdrant Configuration
   QDRANT_HOST=localhost
   QDRANT_PORT=6333

4.4 Running the Application

1. Run the Flask web server:

   flask run

   Or:

   python app.py

   The application will be accessible at http://127.0.0.1:5000.

2. Run other scripts (if applicable):

   • To run the Selenium data crawler:

     python run_selenium_crawler.py

   • To retrain the CNN model:

     python train_model.py

5. DEVELOPMENT TEAM

The project was executed by the STACKLIFO TEAM:

Name	Role
Trần Thế Anh	AI Engineer
Nguyễn Văn Bảo	Developer
Võ Duy Khang	Developer

5. PRODUCT IMAGES

1. LOGIN AND REGISTER:

   

2. HOME PAGE:

3. CHAT BOT:

5. GG MAP:

Report link ---> PRESS ME