ICAR 2025 Workshop - Synthetic Data Generation with Isaac Sim

A tutorial project demonstrating synthetic data generation for robotics using NVIDIA Isaac Sim. This workshop provides tools and examples for generating randomized scenes with a UR10e robot, camera pose randomization, and automated data capture for computer vision and machine learning applications.

📋 Prerequisites

• Docker and Docker Compose (latest versions recommended)
• NVIDIA GPU with compatible drivers (for Isaac Sim)
• Nvidia Container Toolkit to get access to the GPU from the containerized tool
• Python 3.10+ for local development
• Sufficient disk space for Isaac Sim Docker image (~10GB)

🚀 Quick Start

Installation

1. Clone the repository:

   git clone https://github.com/Ekumen-OS/icar_2025_tutorial.git
   cd icar_2025_tutorial

2. Source the workspace and build the Docker image:

   source setup.bash
   icar_sim build

Basic Usage

Run a simple data generation session:

icar_sim run --count 10 -o ./output

This will generate 10 randomized scenes and save the captured data to the ./output directory.

Advanced Usage

Run with custom configuration and additional features:

icar_sim run \
  --count 100 \
  --output-path ./dataset \
  --config sample_config \
  --headless

📖 Usage

Command-Line Interface

The icar_sim command provides a convenient interface for building and running simulations.

Build Command

icar_sim build

Builds the Docker image with all necessary dependencies.

Run Command

icar_sim run [OPTIONS]

Required Arguments:

• --count, -c <NUMBER>: Number of scenes/captures to generate
• --output-path, -o <PATH>: Directory path for captured data

Optional Arguments:

• --headless: Run without GUI (faster, recommended for large datasets)
• --build: Build Docker image before running
• --keepalive: Keep simulation running after completion (useful for debugging)
• --cosmos: Enable CosmosWriter for additional data output formats
• --config <NAME>: Use a specific configuration file from config/ directory (without .yaml extension)
  • Default: sample_config

Configuration Files

Configuration files are YAML-based and located in the config/ directory. They control scene randomization parameters.

Example configuration (config/sample_config.yaml):

camera:
  position: [0, 0, 3.5]
  rotation: [0, 0, 0]
  resolution: [1024, 768]
  pose_constraints:
    look_at: [0, 0, 0]
    x_min: -4.0
    x_max: 4.0
    y_min: -4.0
    y_max: 4.0
    z_min: 0.0
    z_max: 2.0
    camera_distance: [1.0, 4.5]
    camera_altitude: [0.0, 60.0]
    camera_azimuth: [0.0, 360.0]

props:
  num_props: 8
  position_range:
    x: [-0.75, 0.75]
    y: [-0.75, 0.75]
    z: [0.0, 0.0]
  rotation_range:
    z: [0.0, 360.0]

To use a custom configuration:

icar_sim run --count 50 -o ./output --config my_custom_config

🏗️ Project Structure

icar_2025_tutorial/
├── assets/                    # USD assets and 3D models
├── config/                    # YAML configuration files
├── docker/                    # Docker configuration
├── icar_sim_sdg/             # Main Python package
│   ├── randomizers/          # Scene randomization modules
│   ├── writers/              # Data output writers
│   └── entrypoint.py         # Simulation entry point
├── docs/                      # Documentation
├── icar_sim                   # Main CLI script
├── setup.bash                 # Environment setup script
├── CONTRIBUTING.md            # Contribution guidelines
└── LICENSE                    # BSD 3-Clause License

🔧 Development

Setting Up Development Environment

1. Install pre-commit hooks:

   pip install pre-commit
   pre-commit install

2. Run code quality checks:

   pre-commit run --all-files

Code Style

This project uses:

• ruff for linting and formatting
• mypy for type checking
• pre-commit hooks for automated checks

All code must pass these checks before being committed.

Adding Custom Randomizers

To create a custom randomizer:

1. Create a new file in icar_sim_sdg/randomizers/
2. Inherit from the base randomizer pattern
3. Implement the randomize() method
4. Register it in entrypoint.py

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

📄 License

This project is licensed under the BSD 3-Clause License - see the LICENSE file for details.

📬 Contact & Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions