Video coverage
Model accuracy
Plant downtime saved per day
To strengthen the overall business model of the Coca-Cola bottling refranchising in North America, the six largest Coca-Cola Bottlers announced the formation of an Information Technology services company, Coke One North America (CONA) in 2016. CONA runs huge supply chain operations and performs accounting of inventory rigorously. In fact, their warehouses undergo daily and weekly inventory counting processes which are entirely manual and require the warehouse to be inoperative for 6 to 10 hours. Workers must manually identify and count the number of missing stock keeping units (SKUs), scannable printed bar codes, that allow vendors to automatically track the movement of inventory in the warehouse.
However, the movement of SKUs from the warehouse and into trucks is often prone to errors due to manual dependency on scanning the labels; impacting shipment fulfillment that may result in loss of material, poor productivity, and delayed operations. CONA partnered with Quantiphi to leverage machine learning to identify the bins in the stock transport order (STO) inventory of the warehouse, visible from the camera mounted on top of a forklift, and reflect an estimated count of missing pallets in the bin.
Python
TensorFlow
AWS S3
AWS EC2
Quantiphi developed a custom computer vision framework for validating order and shipment accuracy via automated tracking and flagging of incorrect shipments; reporting results on a web user interface in the form of alerts for quick corrective action. This automated inventory count process prevents downtime, errors, and overall operational cost.
