https://hgpu.org/?p=6316
The MOPED framework: Object recognition and pose estimation for manipulation