Object 6-DoF Pose

Key Insight

Knowing an object is "a mug" is not enough to grasp it — the robot needs its full 6-DoF pose: where it sits and how it is turned. 6-DoF pose estimation trains a network to predict that pose directly from an image or point cloud, and fine-tuning it on your own small object set is far cheaper than gathering enough data to train one from scratch. Success is scored with ADD-S, the average 3D distance between the object's points under the predicted pose versus the true pose — computed with a nearest-point match so that symmetric objects (a featureless cylinder looks identical when flipped) are not unfairly penalized for a "wrong" rotation that is actually indistinguishable.