How thed2lab/adaptlil works
Main System Components
a) Java Backend (src/adaptlil/)
• Eye Tracking Integration (gazepoint/ directory)
• Adaptation Engine (adaptations/ directory)
• WebSocket Server (websocket/ directory)
• HTTP Server (http/ directory)
b) Python Server (python/)
• Deep Learning Models
• Flask Server
• Model Training & Evaluation
c) Frontend Visualization (resources/visualization_web/)
• Web Interface
• D3.js Visualizations
• WebSocket Client
Key Relationships to Highlight
a) Data Flow:
• Eye Tracker → Java Backend (Gazepoint API)
• Java Backend → Python Server (Deep Learning inference)
• Python Server → Java Backend (Adaptation decisions)
• Java Backend → Frontend (WebSocket adaptations)
b) Control Flow:
• Frontend ↔ Java Backend (WebSocket communication)
• Java Backend ↔ Python Server (HTTP/REST communication)
Detailed Component Breakdown
Java Backend:
• GazepointSocket
• AdaptationMediator
• VisualizationWebsocket
• Data Processing Pipeline
• Buffer Management
Python Server:
• Flask API
• Model Loading
• Inference Engine
• Deep Learning Models
• LSTM
• Transformer
Frontend:
• Visualization Components
• WebSocket Client
• Adaptation Renderer
• User Interface
Technical Specifications
A technical specifications section showing:
• Java ≥ 11
• Python ≥ 3.9, < 3.11
• CUDA 11+
• WebSocket Protocol
• HTTP/REST APIs
Additional Elements
Callouts for:
• Scalability points
• Error handling paths
• Configuration management
• Model loading/switching capability
Remember to highlight the system's unique features:
• Real-time gaze processing
• Adaptive visualization changes
• Deep learning integration
• Multiple adaptation types
• WebSocket-based updates
This will create a comprehensive yet clear representation of the AdaptLIL system architecture.