Logistics optimization with computer vision in warehouses

The challenge

Counting and checking items in warehouses becomes especially complex when there are disorganized shelves, partially visible products, open boxes, overlapping objects, and lighting variations. Under these conditions, manual inspections are time-consuming and subject to inconsistencies.

The solution developed

CISP1 developed a computer vision application that analyzes images and videos captured in the environment, identifies visible products or SKUs, and consolidates the count across different frames and angles.

The design can combine detection, segmentation, tracking, and multi-frame fusion to reduce duplication and improve counting quality. Low-confidence cases can be directed to human review before the inventory process update.

How the application works

• Captures images or videos of storage areas.
• Detects and segments visible products or SKUs.
• Tracks objects between frames to reduce double counting.
• Consolidates results and flags low-confidence situations.
• Makes the count available for review or integration into the inventory process.

Value for the operation

• Reduces manual visual inventory effort.
• Increases counting consistency in unstructured environments.
• Enables assisted review of low-confidence cases.
• Enables more frequent data generation for inventory management.

Capabilities applied