Glossary

Term Definition Why it matters
Sample An object and its variations that the K-Pick model can be trained on Feeds the K‑Pick training pipeline so the model “knows” each item
View / Texture One camera snapshot of a sample, either depth+color (view) or high‑res RGB (texture) Builds accurate 2.5D and texture models
Point Cloud A collection of 3D points captured by the camera The raw depth data used to generate 3D object models and identify them during execution
Instance Segmentation Model Neural network that detects and outlines each object instance in the scene Core of both U‑Pick (pre‑trained) and K‑Pick (user‑trained) flows
Pick Box User‑defined 3D container in the robot's workspace Restricts detection area and enables collision checks
Environment Active set of one or more pick boxes Defines safe work‑cell limits for planning
Tool Collision Box 3D envelope around the gripper/end‑effector Prevents crashes during approach and retraction
Tool Center Point (TCP) The reference point on the end‑effector from which all robot motions are computed Ensures consistent, accurate pick & place movements
Calibration Computation of the transform that maps camera coordinates into robot coordinates (eye‑in‑hand or hand‑eye) Guarantees that visual data aligns with robot motions
Alignment Method to match sample models to scene data: none, texture‑based, or geometry‑based Determines how pick points orient to the actual objects
Picking Point Exact 3D coordinate on a model where the robot grips Directs the robot to the precise grasp location
Pose A 6‑DOF description (position + orientation) of the robot's end‑effector Specifies the coordinates and orientations of a point in space
Revolution Points Evenly spaced pick points around a symmetric object Automates point placement for objects like cylinders or rings
Category Logical group of samples sharing the same place action Simplifies segmentation classes and downstream application logic
Negative Category Category of samples meant to be ignored by the segmentation model Improves model robustness by teaching it what not to pick
Application (U‑Pick / K‑Pick) A complete workflow: environment, calibration, categories, models, pick points & trained model Encapsulates everything needed to train and run a pick‑place task
URCap Robot‑controller plugin package (e.g., for Universal Robots) Bridges Pick[+] server with the robot's real‑time control system
Digital Output (DO) Robot I/O line used to control peripherals (e.g., lighting) Lets the robot trigger lights or other hardware
Logs Runtime records from the client and server Crucial for debugging errors or unexpected behavior