AI-powered facial health analysis web app — captures a live webcam frame and analyzes skin condition, eye droopiness, and signs of intoxication using MediaPipe FaceMesh and OpenCV.

VitalGaze is a full-stack facial health analysis application. It captures a live webcam frame through the browser, runs it through a computer vision pipeline powered by MediaPipe FaceMesh and OpenCV, and returns a structured health report covering skin redness, pimple detection, eye droopiness, and intoxication indicators.

No cloud APIs. No deep learning models to download. Everything runs locally on your machine.

• 📸 Live webcam capture — grabs a frame directly from the browser via Base64 encoding
• 🗺️ 468-point face mesh — MediaPipe FaceMesh maps precise facial landmarks on the image
• 🔴 Skin redness analysis — measures saturation levels in HSV colour space across the face region
• 🔵 Pimple detection — adaptive thresholding + contour circularity filter to identify acne
• 👁️ Eye droopiness (EAR) — calculates Eye Aspect Ratio for both eyes to detect drooping
• 🚨 Intoxication detection — flags combined high redness + low EAR as intoxication signal
• ⚠️ Abnormality detection — overall health flag based on multi-signal threshold logic
• 📊 Structured JSON output — results rendered cleanly on a dedicated output page
• 🌐 Multi-page Flask UI — Home, About, Docs, Login, Signup, Project pages

Browser captures webcam frame
→ Encodes as Base64 string
→ POST to /results
        │
        ▼
Flask saves decoded image
to Backend/uploaded_img/
        │
        ▼
GET /output triggers
subprocess: python main.py <image_path>
        │
        ▼
FacialHealthAnalyzer.analyze_face()
        │
        ├── MediaPipe FaceMesh
        │   └── Extract 468 facial landmarks
        │
        ├── create_face_mask()
        │   └── Convex hull mask over face region
        │
        ├── analyze_skin()
        │   ├── HSV saturation → redness_level
        │   └── Adaptive threshold + circularity → pimples_detected
        │
        ├── analyze_eyes()
        │   └── Eye Aspect Ratio (EAR) → eye_ratio
        │
        ├── detect_toxication()
        │   └── redness > 180 AND eye_ratio < 0.2 → True/False
        │
        └── determine_abnormality()
            └── Any threshold breach → True/False
        │
        ▼
JSON result → rendered on output.html

• Python 3.9+
• Webcam

# 1. Clone the repository
git clone https://github.com/Corerishi/VitalGaze.git
cd VitalGaze

cd Backend
pip install -r requirements.txt

cd Frontend
python app.py

Visit http://localhost:5000 in your browser.

VitalGaze/
├── Backend/
│   ├── main.py                    # FacialHealthAnalyzer — core CV + ML logic
│   ├── download_landmarks.py      # MediaPipe landmark utilities
│   ├── alcohol_detection_model.pkl
│   ├── dataset_features.joblib
│   ├── uploaded_img/              # Temp storage for captured frames
│   └── requirements.txt
│
├── Frontend/
│   ├── app.py                     # Flask app — routes + webcam frame handler
│   ├── templates/                 # HTML pages (index, about, docs, output, etc.)
│   └── static/                    # CSS, JS, assets
│
└── README.md

• MediaPipe FaceMesh — real-time 468-point 3D facial landmark detection
• Eye Aspect Ratio (EAR) — geometric measure of eye openness used in drowsiness detection
• HSV colour analysis — skin redness measurement using saturation channel
• Adaptive thresholding — pimple detection via contour shape and circularity
• Flask full-stack architecture — frontend-backend separation with subprocess communication
• Base64 image transport — browser webcam capture → server-side OpenCV processing

Rishi Raj
MCA — CHRIST (Deemed to be University)
LinkedIn · GitHub

This project is licensed under the MIT License.