Computer Vision driven return package processing
AI & Data
Semester programme:Artificial Intelligence
Research group:Sustainable Data & AI Application
Project group members:Alexandru Tcaciuc
Arman Parsaravesh
Antonia Dimitrova
Mate Major
Project description
How can a computer vision model be developed to reliably support the unpackaging process by detecting boxes, identifying main components, and reading barcodes in real time?
Context
Logicall, together with its division Drake & Farrell, manages complex logistics operations across Europe and beyond, with a strong focus on reverse logistics, refurbishment, and reuse. One of their key challenges lies in handling the large number of uncontrolled product returns.
Each month, over 150,000 returned packages arrive, which must be unpacked, registered, and processed within two days. This task is highly labor-intensive, requiring skilled workers to manually open packages, capture serial numbers, and match returned components to the correct Return Merchandise Authorization (RMA) numbers before sorting parts for recycling or reuse.
The Dekit & Harvesting Tracking project was initiated to address this challenge through automation and artificial intelligence. By leveraging computer vision, barcode scanning, and object tracking models, the project aims to replace much of the manual effort with an AI-driven solution. The system must process up to 10,000 units per day (roughly one every three seconds), reduce full-time employee (FTE) requirements by half, and decrease training time for new operators from one month to a single day.
Results
This product has a nature of growing as the company grows. The current state provides the client with research that can be applied to a current unpacking setup but has to be updated with new data and settings.