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

Abstract

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: generating, using a fingerprinting network associated with a neural network architecture, feature embeddings corresponding to images; receiving, at an electronic platform, a first non-compliant image; identifying first non-compliant content in the first non-compliant image; and identifying, using multi-modal search options and codes based on the feature embeddings, one or more of the images comprising non-compliant content corresponding to the first non-compliant content in the first non-compliant image. Other embodiments are disclosed herein.

What is claimed is: 
     
         1 . 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:
 generating, using a fingerprinting network associated with a neural network architecture, feature embeddings corresponding to images; 
 receiving, at an electronic platform, a first non-compliant image; 
 identifying first non-compliant content in the first non-compliant image; and 
 identifying, using multi-modal search options and codes based on the feature embeddings, one or more of the images comprising non-compliant content corresponding to the first non-compliant content in the first non-compliant image. 
   
     
     
         2 . The system of  claim 1 , wherein generating the feature embeddings further comprises using a feature extraction component to generate the feature embeddings corresponding to features in the images to preserve feature information of the features in the images. 
     
     
         3 . The system of  claim 1 , wherein the computing instructions, when executed on the one or more processors, further cause the one or more processors to perform an operation comprising:
 searching the codes in one or more databases to identify the one or more of the images by:
 generating, using the fingerprinting network, a feature embedding corresponding to a non-compliant feature of a non-compliant image; 
 generating, using the fingerprinting network, the codes for the images, wherein the codes are based on the feature embeddings; and 
 generating, using the fingerprinting network, a sub-code of the codes for the non-compliant image, wherein the sub-code of the codes is based on the feature embedding corresponding to the non-compliant feature of the non-compliant image, 
 wherein:
 the non-compliant content comprises inappropriate content that is determined to be at least one of: offensive, crude, hateful, or distasteful; 
 the electronic platform comprises a website hosted by the electronic platform and configured to display products and services; and 
 the images comprise images of the products and the services. 
 
   
     
     
         4 . The system of  claim 3 , wherein comparing the sub-code of the codes stored in the one or more databases to identify the one or more of the images comprising the non-compliant content included in the non-compliant image. 
     
     
         5 . The system of  claim 4 , wherein the codes are derived at least in part from the feature embeddings using a hyperplane hashing function. 
     
     
         6 . The system of  claim 5 , wherein the hyperplane hashing function is configured to perform:
 dividing embedding spaces associated with the feature embeddings by using hyperplanes; and   generating the codes corresponding to the hyperplanes using a locality searching hashing function.   
     
     
         7 . The system of  claim 6 , wherein an indexing function uses the codes to facilitate searches to be performed by a search engine. 
     
     
         8 . The system of  claim 1 , wherein the computing instructions, when executed on the one or more processors, further cause the one or more processors to perform an operation comprising:
 providing, using a search component associated with the electronic platform, one or more search interfaces for searching the codes stored in one or more databases to identify the one or more of the images including the non-compliant content similar to, or same as, the non-compliant content included in a non-compliant image.   
     
     
         9 . The system of  claim 8 , wherein providing the one or more search interfaces further comprises:
 providing a seed image that comprises the non-compliant content similar to, or same as, the non-compliant content included in the non-compliant image;   generating one or more binary codes for the seed image utilized to search for the images comprising the non-compliant content similar to, or same as, the non-compliant content included in the non-compliant image; and   comparing the one or more binary codes associated with the seed image with other binary codes stored in the one or more databases.   
     
     
         10 . The system of  claim 1 , wherein the computing instructions, when executed on the one or more processors, further cause the one or more processors to perform an operation comprising:
 executing one or more corrective functions to manage access to the one or more of the images that comprises the non-compliant content which is similar to, or same as, the non-compliant content included in a non-compliant image.   
     
     
         11 . A method being implemented via execution of computing instructions configured to run on one or more processors and configured to be stored at non-transitory computer-readable media, the method comprising:
 generating, using a fingerprinting network associated with a neural network architecture, feature embeddings corresponding to images;   receiving, at an electronic platform, a first non-compliant image;   identifying first non-compliant content in the first non-compliant image; and   identifying, using multi-modal search options and codes based on the feature embeddings, one or more of the images comprising non-compliant content corresponding to the first non-compliant content in the first non-compliant image.   
     
     
         12 . The method of  claim 11 , wherein generating the feature embeddings further comprises using a feature extraction component to generate the feature embeddings corresponding to features in the images to preserve feature information of the features in the images. 
     
     
         13 . The method of  claim 11  further comprising:
 searching the codes in one or more databases to identify the one or more of the images by:
 generating, using the fingerprinting network, a feature embedding corresponding to a non-compliant feature of a non-compliant image; 
 generating, using the fingerprinting network, the codes for the images, wherein the codes are based on the feature embeddings; and 
 generating, using the fingerprinting network, a sub-code of the codes for the non-compliant image, wherein the sub-code of the codes is based on the feature embedding corresponding to the non-compliant feature of the non-compliant image, 
 wherein:
 the non-compliant content comprises inappropriate content that is determined to be at least one of: offensive, crude, hateful, or distasteful; 
 the electronic platform comprises a website hosted by the electronic platform and configured to display products and services; and 
 the images comprise images of the products and the services. 
 
 
 
     
     
         14 . The method of  claim 13 , wherein comparing the sub-code of the codes stored in the one or more databases is used to identify the one or more of the images comprising the non-compliant content included in the non-compliant image. 
     
     
         15 . The method of  claim 14 , wherein the codes are derived at least in part from the feature embeddings using a hyperplane hashing function. 
     
     
         16 . The method of  claim 15 , wherein the hyperplane hashing function is configured to perform:
 dividing embedding spaces associated with the feature embeddings by using hyperplanes; and   generating the codes corresponding to the hyperplanes using a locality searching hashing function.   
     
     
         17 . The method of  claim 16 , wherein an indexing function uses the codes to facilitate searches to be performed by a search engine. 
     
     
         18 . The method of  claim 11  further comprising:
 providing, using a search component associated with the electronic platform, one or more search interfaces for searching the codes stored in one or more databases to identify the one or more of the images including the non-compliant content similar to, or same as, the non-compliant content included in a non-compliant image. 
 
     
     
         19 . The method of  claim 18 , wherein providing the one or more search interfaces further comprises:
 providing a seed image that comprises the non-compliant content similar to, or same as, the non-compliant content included in the non-compliant image;   generating one or more binary codes for the seed image utilized to search for the images comprising the non-compliant content similar to, or same as, the non-compliant content included in the non-compliant image; and   comparing the one or more binary codes associated with the seed image with other binary codes stored in the one or more databases.   
     
     
         20 . The method of  claim 11  further comprising:
 executing one or more corrective functions to manage access to the one or more of the images that comprises the non-compliant content which is similar to, or same as, the non-compliant content included in a non-compliant image.