System requirements
For the fuller explanation of why these matter, see choose and set up your cameras and set up GPU acceleration. This page is the terse version.
Minimum
| Requirement | Detail |
|---|---|
| Camera | One, even a laptop's built-in camera. Gives 2D tracking only, not reliable 3D. |
| Python | A recent Python, installed via a virtual environment manager (Anaconda recommended) |
| OS | Windows, macOS, or Linux |
| GPU | None required, CPU inference works everywhere |
For reliable 3D data
| Requirement | Detail |
|---|---|
| Cameras | 2 minimum, 3+ recommended, connected directly to USB ports (not through a hub) |
| Calibration board | A printed ChArUco board, see why calibration matters |
Optional
| Software | For |
|---|---|
| Blender | Generating an animated 3D scene from your recording |
| VS Code or JupyterLab | Working with the Jupyter notebooks generated per recording |
| A supported GPU | Faster pose estimation, see set up GPU acceleration for which install to use |
GPU support by platform
| Platform | GPU | What you get |
|---|---|---|
| Windows / Linux | NVIDIA | CUDA, or TensorRT for the fastest option |
| Windows | AMD, Intel, or any DirectX 12 GPU | DirectML |
| macOS | Apple Silicon (M1 and newer) | CoreML, automatic |
| Any | None | CPU fallback, always available |