Real-Time Camera Feed Analytics for Retail Shrinkage Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Retail environments face significant shrinkage issues due to mis-scans or no-scans at checkout counters and other points, leading to substantial financial losses, necessitating a robust and accurate system to track consumer-product interactions and detect activities such as picking, placing, scanning, and bagging items.
Innovation Solution
A system utilizing a distributed on-device-AI model with multiple unsupervised deep neural networks processes real-time camera feeds to identify and track objects and activities, including product objects, person objects, and anatomical parts, generating indices for product objects, and classifying activities like scans, no-scans, mis-scans, or thefts without requiring transaction logs or external devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional checkout monitoring methods are used, then system complexity is low, but measurement precision of consumer-product interactions is insufficient
Solution Approach 1:
The system segments the monitoring task into multiple specialized deep neural networks, each responsible for specific detection functions (object detection, activity recognition, anomaly detection). This segmentation allows each component to be optimized for its specific function while maintaining overall system manageability and accuracy.
Solution Approach 2:
The patent replaces traditional mechanical monitoring systems with AI-based deep neural networks that process video feeds. This substitution enables sophisticated analysis of consumer-product interactions without requiring complex physical sensor arrays or manual monitoring infrastructure.
2Measurement precision
If multiple deep neural networks are deployed for accurate activity classification, then measurement precision improves, but use of energy increases
Solution Approach 1:
The system performs preliminary processing by extracting key features and generating intermediate representations before final classification. Multiple DNNs work in a staged manner, with earlier networks preparing data for subsequent networks, reducing redundant computation and energy waste.
Solution Approach 2:
The system employs multiple DNNs that may perform overlapping or redundant analysis, ensuring high accuracy through ensemble methods. While this increases energy consumption, it provides robust activity classification that single models cannot achieve, particularly for detecting subtle shrinkage behaviors.
Data Source
AI summary
A need for a system and method for preventing shrinkage in physical retail store environment using real-time camera feeds is fulfilled in the ongoing description by (a) configuring cameras in a retail store forming a distributed on-device-AI (DODA) model (b) sampling two-dimensional (2D) frames from cameras, (c) generating labels and location for at least one object identified using a first unsupervised deep neural network model (DNN), (d) visually classifying anatomical parts of the human body using a second unsupervised DNN, (e) enlarging and enhancing a product based on labels and location using a third unsupervised DNN, (f) generating an index of product associated with a person using a fourth unsupervised DNN by reidentifying objects from cameras, and (f) automatically classifying activity characterizing the movement of objects including a scan activity, an in-bag activity, a no scan activity, a mis-scan activity, and a theft activity to prevent shrinkage.