Several community members have been trying to get hand-eye calibration working with the Orbecc Gemini 2 camera on Fairino robots (FR5, FR10). Here's a summary of the current state, challenges, and what paths are available.

The Fairino Vision plugin situation

Fairino officially advertises 3D vision integration with the Orbecc Gemini 2:

However, the FR_Vision_Process plugin (FR_Vision_Process_v1.x.x.tar.gz) is extremely difficult to obtain. It's not available on the standard download pages, and even Fairino dealers seem to struggle getting it from the factory.

What works

  • The Orbecc SDK and Orbecc Viewer work fine standalone — you can capture images, point clouds, and depth data without issues
  • The camera hardware is well-supported by Orbecc's own software stack
  • Download: Orbecc Gemini SDK

What doesn't work (easily)

  • The bridge between camera and Fairino controller — there's no standard way to pipe vision data into the robot's motion system
  • Multiple users report being "close but not fully there" with translation errors and mean error values during hand-eye calibration
  • OpenCV-based calibration has been attempted but isn't producing accurate enough results for production use

Available paths for integration

Option 1: Ask your distributor for FR_Vision plugin

This is the official path. The plugin should handle calibration and communication. Whether you can actually get it depends on your relationship with Fairino and your distributor.

Option 2: ROS / ROS2 middleware

The community consensus is that ROS/ROS2 is the most viable path for those who need it working now. Fairino has a ROS2 package available, though it defaults to simulation mode — you need to edit ros2_control.xacro to point to the hardware interface.

Note: MoveIt planning may work but /follow_joint_trajectory might not execute. Check ros2 action list for topic presence.

Option 3: Custom application (Python SDK + OpenCV)

Write your own calibration and vision pipeline using:

  • Orbecc SDK for image capture
  • OpenCV for calibration and image processing
  • Fairino Python SDK for robot positioning

This gives you full control but requires significant development effort.

Option 4: C# SDK + custom application

Some integrators have developed internal applications using the C# SDK with cameras from Sick, Keyence IV4, and similar. These are custom solutions, not publicly available, but prove it can be done.

3D model files for simulation

For integration planning and simulation, Fairino provides STEP files for all robot models:

Downloads — 3D Models

These can be converted to STL/DAE using any CAD software (FreeCAD, Fusion 360, etc.).