Artificial Intelligence (AI)
Artificial Intelligence is a set of algorithms which simulates human-like intelligence. Computers can be programmed to analyze, learn, and understand (and/or solve problems).
Checklens’ AI-technology powers cameras that can to automatically detect any type of product in any store within fractions of a second. The AI instantly recognizes a product when it enters the camera's field of view.
Image Classification
The process of categorizing and assigning a label to a given image.
Checklens AI assigns a category to each recognized product. Products are only recognized, not counted during classification.
In this way, Checklens’ Solutions are able to assign items to their categories, e.g., when recognizing fruits and vegetables in Checkfresh.
We define three different types of image classification:
Binary Classification
In effect when we want to predict if an image belongs to a particular category or not, e.g., the AI technology tells us if there is a product on the image or not, without identifying the particular type of product
Multi-Class Classification
Output of a specific class among a set of available options. E.g., Checklens AI tells us the product’s ID from a range of products in a store.
Multi-Label Classification
Assignment of a set of labels to an image which might not be mutually exclusive, e.g., Checklens AI tells us all the colors that are seen on an image or tells us if a main dish or a side dish on a plate is seen.
Object Detection
The process of estimating the location of objects in an image and identifying their classes.
Checklens AI combines the localization with the classification of a recognized product. Usually, location is described in terms of the object’s bounding box.
This allows Checklens’ Solutions to analyze discovered items in the first place.
Multi-Object Tracking (MOT)
Estimating trajectories of multiple target objects in a given video sequence.
Checklens AI tracks multiple products that are seen within a video. A model is developed in order to link detections across each frame and to generate object trajectories. Hence, this allows us to automatically track products in each video.
This allows Checklens’ Solutions to detect "non-scans" or "miss-scans" of the tracked items.
Action Recognition
Classification of an activity of one or more subjects based on a video sequence of their actions.
Checklens AI classifies certain activities of a shopper or a group of shoppers, e.g., shoppers standing in front of a shelf, shoppers taking products off a shelf, a customer scanning products at a self-checkout, shoppers passing products to other shoppers or customers placing products into a shopping basket, back onto the shelf etc.
This allows Checklens’ Solutions to use customer behavior in order to analyze "non-scans" or "miss-scans."
Face and Other Personal Data Anonymization
Removal of all personal identifiers (e.g., eyes, ears, the entire face) and private or confidential information (e.g., credit card details) from raw data in order to keep all shoppers anonymous throughout the entire store visit.
Checklens AI uses an automatic anonymization system. Regions containing facial information and visible personal/sensitive data are automatically replaced with blurred regions. Therefore personal data is neither recognizable for human intelligence nor for artificial recognition systems.
