US 12,469,255 B2
Systems and methods for identifying different product identifiers that correspond to the same product
Ashlin Ghosh, Ernakulam (IN); Feiyun Zhu, Allen, TX (US); Avinash M. Jade, Bangalore (IN); Lingfeng Zhang, Dallas, TX (US); Amit Jhunjhunwala, Bangalore (IN); Raghava Balusu, Achanta (IN); William Craig Robinson, Jr., Centerton, AR (US); Benjamin R. Ellison, San Francisco, CA (US); Srinivas Muktevi, Bengaluru (IN); and Zhaoliang Duan, Frisco, TX (US)
Assigned to Walmart Apollo, LLC, Bentonville, AR (US)
Filed by Walmart Apollo, LLC, Bentonville, AR (US)
Filed on Feb. 13, 2023, as Appl. No. 18/168,198.
Prior Publication US 2024/0273863 A1, Aug. 15, 2024
Int. Cl. G06V 10/74 (2022.01); G06Q 10/087 (2023.01); G06V 10/762 (2022.01)
CPC G06V 10/762 (2022.01) [G06Q 10/087 (2013.01); G06V 10/761 (2022.01); G06V 2201/07 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A system for processing captured images of objects at a product storage facility, the system comprising:
one or more image capture devices configured to capture images of the objects at the product storage facility;
a trained machine learning model stored in a memory, wherein the trained machine learning model is trained on the images and metadata associated with a plurality of product identifiers, a plurality of product reference images, a plurality of planogram data, or combination thereof; and
a control circuit executing the trained machine learning model configured to:
convert the captured images into a plurality tensors;
group the plurality of product identifiers into one or more clusters based on at least one of visual similarity of corresponding images, textual similarity of corresponding associated descriptions, and associated relationships between product identifiers of the plurality of product identifiers, wherein the visual similarity of the corresponding images is determined based on a calculation of hamming distances pairwise among the plurality of tensors;
determine clusters having common elements that are at least within a similarity threshold of each other;
merge the clusters with the common elements; and
generate a mapping dataset used to retrain the trained machine learning model to identify a plurality of objects, wherein the mapping dataset comprises a plurality of associations of associated product identifiers to a single object.