Instant AI Skin Analysis for Early Detection. Tu primera opinión médica en segundos, no en meses.

🌐 Live Demo: www.dermaskinai.tech 🌐 Devpost devpost.com

DermaSkinAI is a hybrid AI web application designed to act as a preliminary triage tool for skin lesions.

I come from Antofagasta, Chile, a region known for the Atacama Desert and having some of the highest solar radiation levels on Earth. In my hometown, skin cancer is a daily reality, yet getting an appointment with a dermatologist can take 3 to 6 months.

We built DermaSkinAI to bridge this gap: reducing patient anxiety and providing immediate, actionable guidance while they wait for a specialist.

Unlike standard classifiers that just output a label, we use a Hybrid AI Approach:

1. The Eye (Computer Vision): A TensorFlow/Keras model (based on MobileNetV2) analyzes the lesion's visual patterns to detect 7 common pathologies (Melanoma, Nevus, etc.).
2. The Brain (LLM Reasoning): The raw probability data is sent to Google Gemini.
3. The Output: Gemini acts as a "Medical Translator," generating an empathetic, easy-to-understand explanation with triage recommendations.

graph LR
    A[📷 User Photo] -->|Upload| B(🐍 Flask Server)
    subgraph "Hybrid AI Engine"
    B -->|1. Vision Analysis| C{👁️ TensorFlow}
    C -->|Class Probabilities| D[🧠 Google Gemini]
    D -->|Medical Context & Triage| D
    end
    D -->|2. Final Report| E[📱 Client App]
    style C fill:#ff9900,stroke:#333,stroke-width:2px
    style D fill:#4285F4,stroke:#333,stroke-width:2px,color:white