| # 👁️ InsightUX Browser |
|
|
| > AI-powered eye-tracking research browser for usability analysis and user experience evaluation. |
|
|
| InsightUX Browser is a desktop research browser built using **PyWebView**, **MediaPipe**, and an **ONNX-based gaze estimation model**. It enables researchers to record users' gaze behavior while browsing real websites and automatically generates detailed visual analytics including attention heatmaps, dwell timelines, and ranked Areas of Interest (AOIs). |
|
|
| --- |
|
|
| ## ✨ Features |
|
|
| - 👁️ Real-time gaze estimation using a CNN (ONNX Runtime) |
| - 🌐 Integrated research browser built with PyWebView |
| - 🎯 Automatic AOI (Area of Interest) detection |
| - 📊 Attention ranking and dwell-time analytics |
| - 🔥 Screenshot-based heatmap generation |
| - 📈 Timeline visualization of user attention |
| - 💾 Session logging for gaze and DOM interactions |
| - 📷 Automatic screenshot capture during browsing |
| - 🧠 Head pose compensation and illumination normalization |
| - ⚡ Lightweight desktop application |
|
|
| --- |
|
|
| ## Project Structure |
|
|
| ```text |
| InsightUX/ |
| │ |
| ├── browser_session.py # Main browser application |
| ├── calibrate.py # User calibration |
| ├── validate.py # Calibration validation |
| ├── analysis.py # Analytics & report generation |
| ├── inference_pipeline.py # ONNX inference pipeline |
| ├── session_logger.py # Session logging |
| │ |
| ├── models/ |
| │ └── gaze_cnn_v4.onnx |
| │ |
| ├── preprocessing/ |
| │ └── preprocessing_pipeline.py |
| │ |
| ├── sessions/ |
| │ |
| └── requirements.txt |
| ``` |
|
|
| --- |
|
|
| ## Requirements |
|
|
| - Python 3.11 |
| - Webcam |
| - Windows 10 / Windows 11 |
|
|
| --- |
|
|
| ## Installation |
|
|
| Create a virtual environment |
|
|
| ```bash |
| python -m venv venv |
| ``` |
|
|
| Activate it |
|
|
| ```bash |
| .\venv\Scripts\Activate.ps1 |
| ``` |
|
|
| Install dependencies |
|
|
| ```bash |
| pip install -r requirements.txt |
| ``` |
|
|
| --- |
|
|
| ## Usage |
|
|
| ### 1. Calibrate |
|
|
| ```bash |
| python calibrate.py |
| ``` |
|
|
| This generates: |
|
|
| - calibration.pkl |
|
|
| --- |
|
|
| ### 2. Validate (Optional) |
|
|
| ```bash |
| python validate.py |
| ``` |
|
|
| --- |
|
|
| ### 3. Launch Browser |
|
|
| ```bash |
| python browser_session.py |
| ``` |
|
|
| Workflow |
|
|
| 1. Open any website |
| 2. Click on the webpage |
| 3. Press **S** to start tracking |
| 4. Browse normally |
| 5. Press **E** to stop tracking |
| 6. Analytics report opens automatically |
|
|
| --- |
|
|
| ## Generated Outputs |
|
|
| Each session generates |
|
|
| - Raw gaze logs |
| - DOM interaction logs |
| - Website screenshots |
| - Attention heatmaps |
| - Ranked AOIs |
| - Dwell timelines |
| - HTML analytics report |
|
|
| --- |
|
|
| ## Technology Stack |
|
|
| - Python |
| - PyWebView |
| - MediaPipe |
| - OpenCV |
| - NumPy |
| - ONNX Runtime |
| - JavaScript |
| - HTML/CSS |
|
|
| --- |
|
|
|
|
|
|
|
|
