Walmart Apollo, LLC

Binwei Yang

Cun Mu

Alessandro Magnani

Zheng Yan

Behzad Ahmadi

Omer Ovenc

Jun Zhao

Guang Yang

Theban Stanley

Ari Kast

Xinyi Liu

Manik Goyal

Rashad Eletreby

A system including one or more processors and one or more non-transitory computer-readable storage devices storing computing instructions that, when run on the one or more processors, cause the one or more processors to execute operations: storing image hashes corresponding to images for items offered via an electronic platform; receiving, from a computing device, a search query via a search engine included on the electronic platform; executing a relevance function; executing a visual diversity ranking function that reorders the relevant set of search results to generate a ranked set of search results, wherein the visual diversity ranking function utilizes the image hashes corresponding to the subset of the items to reorder the relevant set of search results; and transmitting the ranked set of search results to the computing device. Other embodiments are disclosed herein.

Systems and methods for improving visual diversities of search results in real-time systems with large-scale databases

A method including automatically determining, by a machine learning model trained based at least in part on sample items stored in a sample database, a query embedding vector for a query image of a query item. The method further can include determining, based on a respective embedding distance between the query image of the query item and a respective image of each of the sample items, neighboring items from among the sample items. The respective embedding distance can be calculated based on the query embedding vector for the query image and a respective embedding vector for the respective image of each of the sample items. Each of the sample items can include the respective image and at least one respective item label. The method also can include determining a respective normalized weight for each of the neighboring items based on the respective embedding distance between the query image and the respective image of the each of the neighboring items. The method additionally can include determining a query item label of the query item based on a weighted majority vote by the neighboring items via the respective normalized weight for the each of the neighboring items. The method further can include upon determining that the query item label of the query item is different from a first item label of the query item. storing the query item with the query item label in a product database. The method also can include selectively updating the sample items stored in the sample database from items in the product database. In addition, the method can include re-training the machine learning model based at least in part on the sample items in the sample database, as updated. Other embodiments are disclosed.

System and method for determining item labels based on item images

A system comprising one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform operations comprising: training a visual similarity model to output one or more vector representations in a vector space; determining one or more nearest neighbors based on embeddings nearest to a respective embedding in the vector space for the item; ranking the one or more nearest neighbors for the item in an order; re-ranking the order of the one or more nearest neighbors of the item into a re-ranked order; inserting, into a display of the item, the one or more nearest neighbors of the item; and facilitating altering a graphical user interface (GUI) to convey the display of the item. Other embodiments are disclosed herein.

Systems and methods for retraining of machine learned systems

A method including pre-screening second images from a database for a search result based on a first image and also based on respective substring distances between first binary substrings for the first image and respective second binary substrings for each of the second images. The method further can include after the pre-screening, determining a respective image distance for each third image of third images of the search result, wherein the respective image distance is between the first image and the each third image of the third images of the search result, and after determining the respective image distance for each third image of the third images of the search result, when the respective image distance for a particular one of the third images is greater than a predetermined image distance threshold, removing the particular one of the third images from the search result before the search result is displayed on a user device requesting the search result. Other embodiments are disclosed.

Image searching using a full-text search engine

A method including determining neighboring items from among sample items based on a respective embedding distance between a query image of a query item and a respective image of each of the neighboring items. The each of the sample items can comprise the respective image and at least one respective item label. The method further can include determining a respective normalized weight for each of the neighboring items based on the respective embedding distance between the query image and the respective image of the each of the neighboring items. The method also can include determining a query item label of the query item based on a weighted majority vote by the neighboring items via the respective normalized weight for the each of the neighboring items. Other embodiments are disclosed.

Systems and methods including one or more processors and one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and perform: receiving images at an electronic platform; generating, using a fingerprinting network associated with a neural network architecture, feature embeddings corresponding to the images; generating, using the fingerprinting network, codes based on the feature embeddings; storing the codes in one or more databases; receiving, using a search component associated with the electronic platform, a non-compliant image comprising non-compliant content; and searching the codes in the one or more databases to identify one or more of the images that include non-compliant content similar to, or the same as, the non-compliant content included in the non-compliant image. Other embodiments are disclosed herein.

Systems and methods for identifying non-compliant images using neural network architectures

Systems and methods including one or more processors and one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and perform: training a visual similarity model; determining one or more nearest neighbors for an item using the visual similarity model, as trained; ranking the one or more nearest neighbors using results of the visual similarity model, as trained; re-ranking the one or more nearest neighbors using a learning-to-rank algorithm; and facilitating altering a graphical user interface (GUI) displaying the item based upon the one or more nearest neighbors, as re-ranked. Other embodiments are disclosed herein.

A method including bit-operation and sub-code/substring filtering for image searching using a full-text search engine. The method can include determining a first binary vector comprising first binary substrings for a first image. The method also can include obtaining a respective second binary vector comprising second binary substrings for each of second images from a database. The method additionally can include determining a respective substring distance for each of the binary substring for each of the second images. The respective substring distance can be between at least a pair of a first binary substring of the first binary substrings of the first binary vector and a respective corresponding second binary substring of the second binary substrings of the respective second binary vector for each of the second images. In some embodiments, the method further can include after determining the respective substring distance for each of the binary substring for each of the second images, when the respective substring distance for one or more of the second images is not greater than a predetermined substring distance threshold, including the one or more of the second images in a search result. The method also can include determining a respective image distance for each respective third image of the search result, the respective image distance being between the first image and the each respective third image of the search result. The method additionally can include after determining the respective image distance for the each respective third image of the search result, when the respective image distance is greater than the predetermined image distance threshold, culling the each respective third image from the search result. Other embodiments are disclosed.

Systems and methods including one or more processing modules and one or more non-transitory storage modules storing computing instructions configured to run on the one or more processing modules and perform acts of receiving a plurality of normalization rules, using the plurality of normalization rules and a multi-class classification algorithm to train a classification model for one or more product attributes of a plurality of products, determining additional normalization rules for transforming additional structured values to additional final normalized values based on a classification of un-normalized additional raw source values with the classification model, validating the additional normalization rules, and adding the additional normalization rules to the classification model if the additional normalization rules are validated.

Growing community of inventors

Sunnyvale, CA, United States of America

Binwei Yang