pyRecFusionSDK

Easy multi-sensor mesh reconstruction with Python bindings to RecFusion SDK (C++).

Changes to original API

This project slightly modifies or adds new functionalites to the original RecFusion SDK API:

• RecFusionSDK version string: version() -> str
• Open all available sensors: SensorManager.open_all() -> list[Sensor]
• Open any available sensor: SensorManager.open_any() -> Sensor | None
• Obtain list of sensor color and depth formats:
  • Sensor.color_formats: list[SensorFormat]
  • Sensor.depth_formats: list[SensorFormat]
• Initialize playback recorder using sensor options: RFSRecorder.init_from_sensor(sensor: Sensor) -> None
• Initialize empty ColorImage and DepthImage for a given Sensor (using its resolution etc.):
  • ColorIamge.allocate_for_sensor(sensor: Sensor) -> ColorImage
  • DepthIamge.allocate_for_sensor(sensor: Sensor) -> DepthImage
• Save images to *.png files:
  • ColorImage.to_image(filename: str, compression: int = 3) -> None
  • DepthImage.to_image(filename: str, compression: int = 3) -> None

Usage

Requirements:

• Installed locally the RecFusion SDK

Then:

pip install pyRecFusionSDK

Publish new version to PyPi

1. Remove dist/ and build/
2. Run uv build --sdist
3. Run uv publish
4. Type the token!

Examples

We provide several examples of using RecFusionSDK with Python:

• capture_playback.py - capture sequence of images as playback file (*.rfs)
• capture_png.py - capture single frame to image (*.png)
• machine_token.py - display machine token required for offline license
• reconstruct_from_playback.py - reconstruct mesh from playback file (*.rfs)
• reconstruct_sensor.py - reconstruct mesh directly from sensor data
• reconstruct_sensor_listener.py - (NOT WORKING) reconstruct mesh directly from sensor data using sensor listener
• show_sensor_info.py - show information about sensor (e.g. model, UUID, resolutions)