Method and System for Detecting Drift in Image Streams

Abstract

Methods and systems disclosed herein may quantify a representation of a type of input an image analysis system should expect. The image analysis system may be trained on the type of input the image analysis system should expect using a first image stream. A first model of the type of input that the image analysis system should expect may be built from the first image stream. After the first model is built, a second image, or a second image stream, may be compared to the first model to determine a difference between the second image, or second image stream, and the first image stream. When the difference is greater than or equal to a threshold, a drift may be detected and steps may be taken to determine the cause of the drift.

1 . A method comprising:
 comparing, by a server, a second model, representing a second probability distribution of a second plurality of features appearing in a second stream, to a first model, representing a first probability distribution of a first plurality of features appearing in a first stream;   determining, based on the comparing the second model to the first model, a difference between the second stream and the first stream; and   cause, based on the difference being greater than or equal to a threshold, indication that the second stream differs from the first stream.   
     
     
         2 . The method of  claim 1 , further comprising:
 building, based on the first plurality of features appearing in the first stream, the first model.   
     
     
         3 . The method of  claim 2 , wherein building the first model further comprises:
 determining, based on first data in the first stream, a first numeric representation;   determining, based on second data in the second stream, a second numeric representation;   determining, based on identifying one or more features that appear in both the first numeric representation and the second numeric representation, the first plurality of features appearing in the first stream; and   generating, based on how often each of the first plurality of features appear in both the first numeric representation and the second numeric representation, the first model.   
     
     
         4 . The method of  claim 1 , further comprising:
 building, based on a second stream, the second model.   
     
     
         5 . The method of  claim 1 , comprising:
 generate, via an application, a notification that the second stream differs from the first stream.   
     
     
         6 . The method of  claim 1 , further comprising:
 issuing, via an application, at least one command to an input source to correct a cause of the difference between the second stream and the first stream.   
     
     
         7 . The method of  claim 6 , wherein the input source comprises one or more scanners associated with an automated teller machine. 
     
     
         8 . The method of  claim 1 , wherein:
 the first stream comprises a first plurality of images obtained via an input source; and   the second stream comprises a second plurality of images obtained via the input source.   
     
     
         9 . A computing device comprising:
 one or more processors; and   memory storing instructions that, when executed by the one or more processors, cause the computing device to:
 compare a second model, representing a second probability distribution associated with a second stream, to a first model, representing a first probability distribution associated with a first stream; 
 determine, based on comparing the second model to the first model, a difference between a second plurality of features appearing in the second stream and a first plurality of features appearing in the first stream; and 
 cause, based on the difference being greater than or equal to a threshold, an indication that the second stream differs from the first stream to be displayed. 
   
     
     
         10 . The computing device of  claim 9 , wherein the instructions, when executed by the one or more processors, cause the computing device to:
 generate, via an application and based on the difference being greater than or equal to the threshold, a notification that the second stream differs from the first stream.   
     
     
         11 . The computing device of  claim 9 , wherein the instructions, when executed by the one or more processors, cause the computing device to:
 issue, via an application, at least one command to an input source to correct a cause of the difference between the second stream and the first stream.   
     
     
         12 . The computing device of  claim 9 , wherein the instructions, when executed by the one or more processors, cause the computing device to:
 determine, based on first data in the first stream, a first numeric representation;   determine, based on second data in the second stream, a second numeric representation;   determine, based on identifying one or more features that appear in both the first numeric representation and the second numeric representation, a plurality of features appearing in the first stream; and   generate, based on how often each of the plurality of features appear in both the first numeric representation and the second numeric representation, the first model.   
     
     
         13 . The computing device of  claim 9 , further comprising:
 an interface configured to:
 receive, from an input source, the first stream; and 
 receive, from the input source, the second stream. 
   
     
     
         14 . The computing device of  claim 13 , wherein the input source comprises one or more scanners associated with an automated teller machine. 
     
     
         15 . The computing device of  claim 9 , wherein:
 the first stream comprises a first plurality of images obtained via an input source; and   the second stream comprises a second plurality of images obtained via the input source.   
     
     
         16 . The computing device of  claim 9 , wherein the difference being greater than or equal to the threshold indicates a change between the first stream and the second stream. 
     
     
         17 . A non-transitory computer-readable medium storing instructions that, when executed, cause a computing device to:
 compare a second model, representing a second probability distribution of a second plurality of features appearing in a second stream, to a first model, representing a first probability distribution of a first plurality of features appearing in a first stream;   determine, based on comparing the second model to the first model, a difference between the second stream and the first stream;   cause, based on the difference being greater than or equal to a threshold, an indication that the second stream differs from the first stream to be displayed, wherein the difference being greater than or equal to the threshold indicates a change between the first stream and the second stream; and   generate, via an application, a notification indicating the change between the first stream and the second stream.   
     
     
         18 . The non-transitory computer-readable medium of  claim 17 , wherein the instructions, when executed, cause the computing device to:
 determine, based on first data in the first stream, a first numeric representation;   determine, based on second data in the second stream, a second numeric representation;   determine, based on identifying one or more features that appear in both the first numeric representation and the second numeric representation, a plurality of features appearing in the first stream; and   generate, based on how often each of the plurality of features appear in both the first numeric representation and the second numeric representation, the first model.   
     
     
         19 . The non-transitory computer-readable medium of  claim 17 , wherein:
 the first stream comprises a first plurality of images obtained via an input source; and   the second stream comprises a second plurality of images obtained via the input source.   
     
     
         20 . The non-transitory computer-readable medium of  claim 17 , wherein the instructions, when executed, cause the computing device to:
 issue, via an application, at least one command to an input source to correct a cause of the difference between the second stream and the first stream.