PlateShapez – Tool for generating adversarially perturbed license plate overlays

Summary

📜 PlateShapez A research tool for generating adversarially perturbed license plate overlays on vehicle images, producing structured datasets with reproducibility, transparency, and ethical guardrails. Design Principle: user-first, safe by default, hackable by experts. 🚀 Quick Start 🎬 Try the Demo First! For a complete walkthrough that creates test images and demonstrates both CLI and Python API: # Run the integrated CLI demo (recommended) uv run advplate demo # With automatic cleanup uv run advplate demo --cleanup # Or run the demo script directly uv run python examples/demo_full_workflow.py This demo will: Create synthetic car backgrounds and license plate overlays Show CLI usage with custom configurations Demonstrate Python API with different perturbations Generate datasets and show output structure Display metadata and results analysis Prerequisites uv (Python package manager) installed: # Install uv curl -LsSf https://astral.sh/uv/install.sh | sh Installation # Clone the repository git clone https://github.com/benjordan/plateshapez.git cd plateshapez # Create virtual environment and activate uv venv source .venv/bin/activate # Install dependencies uv sync # or uv sync --group dev for development dependencies # Install the CLI tool uv pip install -e . Basic Usage Prepare your data: project/ ├── backgrounds/ # Vehicle images (JPG) ├── overlays/ # License plate images (PNG with alpha) └── config.yaml # Optional configuration Generate dataset: # Generate dataset with defaults uv run advplate generate # Generate with custom config and seed for reproducibility uv run advplate generate --config my_config.yaml --seed 42 # Preview generation plan uv run advplate generate --dry-run Explore available options: # Run interactive demo uv run advplate demo # List available perturbations uv run advplate list # Show current configuration uv run advplate info --as yaml # Print example configuration uv run advplate examples # Show version uv run advplate version Python API from pl...