Kurly Develops On-device Mis-delivery Detection System Based on Delivery-Completion Photos

·2026.04.28 00:00

Key point

Kurly has developed an On-device system that compares the similarity of delivery-completion photos to detect mis-deliveries in real time.

Details

Kurly's Data Service Development Team unveiled an On-device system that detects mis-deliveries by re-identifying locations (Re-Identification) based on the similarity of delivery-completion photos. Delivery accuracy is directly tied to customer experience and cost, but because mis-deliveries occur infrequently, detecting them with existing methods was difficult.

Ensuring Data Quality

Since delivery-completion photos are taken with managers' smartphones, they can contain abnormal noise. Therefore, a preprocessing step that filters out low-quality images is essential to ensure the validity of similarity comparisons.

computer-vision on-device-ai logistics kurly re-identification

Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Kurly Integrates OMS and TAM to Build an Automated Order Distribution System

Kurly Internalizes Multi-Building Address Refinement to Reduce External API Dependency and Achieve 99.2% Accuracy