System and method for determining out-of-stock products
An automated inventory monitoring system includes an image capture module able to create an image of an aisle of a retail store. Product images and shelf label and peg label images are identified in the image and products are associated with product labels based on the positioning of the products with respect to the labels. Based on the association between labels and products, out-of-stock products are detected and reported to the retail store.
1. A system, comprising:
one or more processors;
one or more sensors coupled to the one or more processors; and software implementing:
a product detection neural network for detecting products in images captured by the one or more sensors;
a label detection neural network for detecting labels in images captured by the one or more sensors;
a product type classifier for identifying products as any one of a shelf product, a peg product, a grill product, or a shelf-ready package;
a shelf ready classifier for determining whether a shelf-ready package is empty or not empty; and
a label type classifier for identifying label types;
wherein the software causes the system to perform the functions of:
obtaining a representation of an aisle containing stocked products from the one or more sensors;
identifying product labels in the representation;
classifying the product labels as specific types of labels using the label type classifier;
identifying products in the representation;
classifying the products as a specific type of product using the product type classifier;
identifying an area associated with each product label, based on the type of the label;
associating products with product labels, based on a placement of the product with respect to a label of the same type as the product; and
identifying, as out-of-stock products, products identified by product labels having no associated products.
2. The system of claim 1 , further comprising:
reporting out-of-stock products.
3. The system of claim 1 , the representation of the aisle comprising an image.
4. The system of claim 3 further comprising:
identifying images of product labels in the image of the aisle; and
identifying images of products in the image of the aisle.
5. The system of claim 1 wherein the label detection neural network identifies the product labels in the representation of the aisle and places a bounding box around each product label; and
wherein the product detection neural network identifies products in the representation of the aisle and places a bounding box around each product, further comprising:
establishing a coordinate system on the representation of the aisle;
determining a size and position of the product label bounding box within the coordinate system; and
determining a size and position of the product bounding box within the coordinate system.
6. The system of claim 5 , wherein the label detection neural network outputs the coordinates of the bounding boxes for each product label with respect to the coordinate system.
7. The system of claim 5 , wherein the product detection neural network outputs the coordinates of the bounding boxes for each product with respect to the coordinate system.
8. The system of claim 1 , the area associated with each product label classified as a shelf label being the section of shelf defined by a product label and an adjoining product label on the same shelf.
9. The system of claim 1 , wherein the type of the product is identified as a shelf-ready packaging product, further comprising:
identifying a product as being out of stock when a shelf-ready packaging is determined to be empty.
10. The system of claim 5 , each bounding box for product labels and products being represented by a tuple of data containing at least coordinates of the bounding box with respect to the coordinate system and the width and height of the bounding box.
11. The system of claim 5 , further comprising:
identifying a location of shelves within the representation of the aisle.
12. The system of claim 11 , the location of shelves in the representation of the aisle being identified by a model trained to output a tuple defining a bounding box for each shelf.
13. The system of claim 11 , the location of the shelves in the representation of the aisle being inferred by identifying when bounding boxes for a plurality of product labels are aligned in a straight line.
14. The system of claim 11 , wherein product labels having a location co-located with a shelf location are associated with shelf products and wherein product labels not co-located with a shelf location are associated with peg products.
15. The system of claim 2 , wherein reporting out-of-stock products further comprises:
obtaining a high-resolution image containing the product label associated with the out-of-stock product; and
reading the identity of the product from the high-resolution image; and
reporting the identity of the out-of-stock product.
16. The system of claim 15 wherein the identity of product is obtained by reading a bar code or QR code identifying the product from the high-resolution image.
17. The system of claim 15 wherein the identity of product is obtained by submitting the product to a classifier.
18. The system of claim 1 wherein the representation of the aisle is collected by a mobile inventory monitoring system that autonomously traverses the aisle.
19. The system of claim 18 wherein the representation of the aisle is a unified representation comprising multiple representations collected by the mobile inventory monitoring camera system at varying vertical and horizontal positions, the multiple representations being stitched together to form the unified representation.
20. The system of claim 3 wherein obtaining a representation of an aisle comprises:
capturing images of sections of the aisle; and
stitching representations for two or more sections together to create a unified image.
21. The system of claim 20 , the unified image being a panoramic image showing the entirety of the aisle.