Auto-enrollment for Computer Vision Recognition System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current POS systems face challenges in accurately identifying items without barcodes, such as fresh produce or uniquely shaped items, requiring manual input or barcode scanning, which can be time-consuming and inefficient.
Innovation Solution
An automated process using machine learning models trained with images captured by cameras at POS systems, allowing for the recognition of new items and verification of existing ones, enabling the system to detect items without relying on barcodes or manual input, and updating the model automatically as new items are added to inventory.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual input or barcode scanning is used to identify items, then item identification can be achieved, but transaction speed and efficiency deteriorate
Solution Approach 1:
The system performs self-enrollment by automatically capturing images of new items, training the ML model, and updating itself without human intervention. The computer vision system automatically identifies items by capturing images and processing them through the ML model, eliminating the need for manual barcode scanning or price look-up operations.
Solution Approach 2:
The patent replaces manual mechanical operations (barcode scanning, manual price look-up, typing) with an automated computer vision system using ML models. The system captures images of items and uses automated image processing and ML inference to identify items, substituting human-operated mechanical processes with automated optical and computational systems.
2Adaptability or versatility
If the ML model is updated to recognize new items, then item recognition capability improves, but system complexity increases
Solution Approach 1:
The system collects and processes images of new items in advance, training the ML model before deployment. By performing preliminary training with captured images and validating the model updates before pushing to production, the system prepares adaptation capabilities ahead of time, reducing the complexity of real-time updates.
Solution Approach 2:
The system implements a feedback loop where the ML model's performance is continuously evaluated using validation datasets. If the model meets accuracy thresholds, updates are pushed to the system; if not, further training iterations occur. This feedback mechanism automates the complexity management by providing clear criteria for model acceptance and update deployment.
3Measurement precision
If the ML model is retrained with new items, then accuracy for new items improves, but accuracy for previously trained items may deteriorate
Solution Approach 1:
Before pushing model updates to production, the system performs preliminary validation by testing the updated model against a validation dataset containing previously trained items. This advance verification ensures that new item recognition improvements do not compromise existing item accuracy, allowing corrective iterations before deployment.
Solution Approach 2:
The system implements a cushioning mechanism by maintaining validation datasets of previously trained items and requiring that model updates meet minimum accuracy thresholds for these items. This protective measure prevents catastrophic forgetting and ensures that improvements in new item recognition do not degrade performance on existing items.
Data Source
AI summary
This disclosure describes an automated process for training an ML model used by a computer vision system in a point of sale (POS) system to recognize a new item. Instead of relying on a manual process performed by a data scientist, the automated process can use images of a new (i.e., unknown) item captured at one or more POS systems to then retrain the ML model to recognize the new item. That is, the images of the item are used to retrain the ML model and to test the accuracy of the updated ML model. If the updated ML model can confidently identify the new item, the updated ML model is then used by the computer vision system to identify items at the POS system.