Computer Vision to Optimize Warehouse Operations
Retail & CPGBusiness Impacts
100% target picking accuracy
<1 second inference time
Saved time and money spent on incorrect shipments
Reduced shortfall in warehouse inventory
Customer Key Facts
- Location : North America
- Industry : Retail & CPG
Leveraging Real-Time Computer Vision to Optimize Warehouse Operations
The customer is a multinational beverage manufacturer whose warehouses ship mixed pallets that comprise different products stacked on a pallet to form one SKU. The pickers manually assemble these SKUs on a pallet jack, making the process error-prone.
These errors occur either in the quantity or flavor of the picked products despite multiple checks in the warehouse. This leads to incorrect shipments to customers, eventually resulting in a loss of revenue. There is no way to flag and prevent these discrepancies at the time of picking.
Challenges
- Setting up hardware with required specs
- Differentiating between products
- Data capture for new products
- Poor internet coverage in the warehouse
Technologies Used
NVIDIA Jetson Xavier NX
NVIDIA Deepstream
Triton Inference Server
Solution
Quantiphi developed a solution to identify the products being picked while assembling a pallet using computer vision and real-time edge inferencing. The video stream of the building of the pallet was processed using NVIDIA® Jetson NX.
An ML pipeline was also created for the detection and classification of objects. It was capable of tracking the quantity of all the products placed on the pallet and validating it with the order details. The real-time alerts were raised on a screen to notify the picker of a mis-pick in product or quantity.
Results
- Enhanced warehouse mixed pallet picking accuracy
- Improved cost and time efficiency by minimizing incorrect shipments
- Optimized warehouse operations
