Vision Based Object Detection & Palletizing with YOLO

Overview

• An automated sorting cell: an external camera feeds YOLO, objects are classified into categories, and the Dobot picks each with suction and places it into the category pallet.
• Closed loop autonomy combining lightweight object detection with code driven motion routines.

Design & Build

• Detection: Ultralytics YOLOv8s classifies workspace objects in real time from the camera feed, gating every pick decision.
• Motion: pydobot motion primitives across four taught poses: home, pickup, and two category drop zones, with suction engagement timed around the vacuum response.
• Loop: Detect, classify, pick, palletize, repeat, with each cycle logged for accuracy and consistency.

Challenges

• Serial communication delays between detections and motion.
• Suction grip consistency across repeated cycles.
• Detection confidence under varying lighting.

Result

• Objects consistently detected, classified, and placed into the correct pallets across repeated trials, with stable cycle times and placement accuracy within a few millimeters.

Skills & Tools