A real-time computer vision system for game interface analysis, designed to extract structured information from on-screen elements and react to visual events with deterministic, configurable logic.
Pixel-Eye focuses on perception, signal extraction, and event-driven responses, treating the game as a black box and relying exclusively on visual input.
Pixel-Eye monitors game windows in real time and processes HUD elements such as health, resources, alerts, and contextual signals.
Core goals:
- Real-time visual perception using screen capture
- Deterministic detection of HUD states and events
- Configurable alerting and response mechanisms
- Clear separation between capture, processing, and reaction layers
The project prioritizes engineering discipline and system clarity over automation gimmicks.
- Screen capture and region-of-interest management
- OpenCV-based image preprocessing
- Deterministic pixel and pattern analysis
- Configurable thresholds and detection rules
- Event-driven alert pipeline (visual/audio hooks)
- OCR-based text extraction for dynamic HUD elements
- Overlay rendering for diagnostics and feedback
- Data recording for post-match analysis
- ML-based detectors for complex visual patterns
pixel-eye/
├── core/ # Capture and processing primitives
│ ├── capture.py
│ ├── processor.py
│ └── detector.py
│
├── modules/ # Domain-specific logic
│ ├── hud_reader.py
│ ├── alerts.py
│ └── analytics.py
│
├── utils/ # Configuration and helpers
│ ├── config.py
│ └── helpers.py
│
├── tests/ # Validation and diagnostics
└── main.py # Application entry point
The architecture enforces strict separation of concerns, allowing individual modules to evolve independently.
- Python 3.8+
- OpenCV — real-time image processing
- NumPy — pixel-level analysis
- Pillow — image manipulation
- PyAutoGUI — window detection and capture
- Tesseract OCR — text extraction (optional)
- PyTorch — ML-based detectors (optional)
- pygame — debug rendering and overlays
git clone https://github.com/your-username/pixel-eye.git
cd pixel-eye
python -m venv venv
source venv/bin/activate # Linux / macOS
venv\Scripts\activate # Windows
pip install -r requirements.txtOptional OCR dependency:
- Windows: https://github.com/UB-Mannheim/tesseract/wiki
- Linux:
sudo apt install tesseract-ocr
from pixel_eye import PixelEye
eye = PixelEye(game_window="Game Name")
eye.set_roi(x=100, y=50, width=800, height=600)
eye.start(
detect_health=True,
health_threshold=30,
enable_overlay=True
)All thresholds, regions, and behaviors are configurable.
- Assistive systems for accessibility
- Automated detection of game states (e.g. end-of-match)
- Competitive analysis and performance feedback
- Vision-based environments for AI research
This project is not intended as:
- A cheat engine
- An input automation framework
- A plug-and-play competitive advantage tool
It is intended as:
- A real-time perception layer for interactive systems
- A reference architecture for vision-driven agents
- A foundation for experimentation in applied computer vision
- ROI auto-calibration
- Robust OCR pipelines
- Overlay customization
- ML-assisted pattern detection
- Cross-game configuration profiles
- External event integration (API/hooks)
This software is provided for educational, research, and accessibility purposes.
Usage must comply with the terms of service of any third-party software it interacts with.
MIT License.