Performing actions based on evaluation and comparison of multiple input processing schemes

Abstract

A computer-implemented method of sensor input processing, implemented by an agricultural platform comprising a processor and a sensor, includes capturing, using the sensor, sensor images of a vicinity of a target object of a time interval during which a treatment is applied to the target object; processing the sensor images using one or more machine learning (ML) algorithms wherein at least one ML algorithm uses an ML model trained to detect a presence of a treatment action in the vicinity of the target object; and providing, selectively based on a result of detecting the presence of the treatment action in the vicinity of the target object, an outcome of the processing for further processing.

1 . A computer-implemented method of sensor input processing, implemented by an agricultural system comprising a processor and one or more sensors, comprising:
 receiving a sensor input from the one or more sensors;   accessing a first image from the sensor input;   applying a first image processing scheme on the first image, wherein the first image processing scheme implements a first machine learning (ML) algorithm configured to detect plant objects or target objects on a portion of the first image;   detecting a first target object in a first portion of the first image, the detection comprising identifying at least a first class associated with the first target object;   accessing the first portion of the first image for further processing;   applying a second image processing scheme on the first portion of the first image, wherein the second image processing scheme implements a second ML algorithm configured to classify a received image, the received image including the first portion of the first image;   identifying at least a second class associated with the first portion of the first image;   comparing results of the first class with the second class; and   performing a subsequent action based on the comparison.   
     
     
         2 . The computer-implemented method of  claim 1 , wherein the first portion of the first image is a tile and the detected first target object in the first portion of the first image is a smaller portion of the first portion, the smaller portion represented by a bounding box associated with pixels of the first image. 
     
     
         3 . The computer-implemented method of  claim 2 , wherein the second ML algorithm classifies the smaller portion of the first portion of the first image. 
     
     
         4 . The computer-implemented method of  claim 2 , wherein the second ML algorithm classifies the smaller portion represented by the bounding box. 
     
     
         5 . The computer-implemented method of  claim 1 , wherein detecting a first target object in the first portion of the first image comprises generating a bounding box associated with the first portion of the first image. 
     
     
         6 . The computer-implemented method of  claim 1 , wherein the first portion of the first image is an image patch. 
     
     
         7 . The computer-implemented method of  claim 6 , wherein identifying at least a second class comprises classifying the image patch. 
     
     
         8 . The computer-implemented method of  claim 1 , wherein the first class and the second class can be classifications from a plurality classifications including a crop class, a weed class, a background class, or a soil class. 
     
     
         9 . The computer-implemented method of  claim 1 , wherein the first ML algorithm is configured to assign more than one class of a plurality of classifications to one or more detected objects, including plant objects and target objects. 
     
     
         10 . The computer-implemented method of  claim 9 , wherein each class, including the first class and second class, has a confidence level associated with the class. 
     
     
         11 . The computer-implemented method of  claim 1 , wherein the second ML algorithm is configured to assign more than one class of a plurality of classifications to one or more detected objects, including plant objects and target objects. 
     
     
         12 . The computer-implemented method of  claim 11 , wherein each class, including the first class and second class, has a confidence level associated with the class. 
     
     
         13 . The computer-implemented method of  claim 1 , wherein each of the first class and second class has a confidence level. 
     
     
         14 . The computer-implemented method of  claim 13 , wherein comparing results of the first class with the second class comprises analyzing and comparing each confidence level of the first class and the second class with each other. 
     
     
         15 . The computer-implemented method of  claim 1 , wherein performing a subsequent action includes performing a treatment action on a real-world object associated with the detected first target object. 
     
     
         16 . The computer-implemented method of  claim 15 , wherein the subsequent action comprises determining to treat the real-world object associated with the detected first target object based on comparing the results of the first class and the second class. 
     
     
         17 . The computer-implemented method of  claim 1 , wherein, upon determining that the first class and the second class are not a match, sending the results of the first class and the second class, including the first portion of the first image to a user interface for further review. 
     
     
         18 . The computer-implemented method of  claim 17 , wherein a selection by a user of the user interface to verify a correct classification of the detected first target object is indexed for training of a first ML model associated with the first ML algorithm. 
     
     
         19 . The computer-implemented method of  claim 1 , wherein the first class is a weed species and the second class is a soil pattern. 
     
     
         20 . A computer-readable medium having code stored thereon, the code, upon execution by a processor, causing the processor to implement a method of sensor input processing, comprising:
 receive a sensor input from the one or more sensors;   access a first image from the sensor input;   apply a first image processing scheme on the first image, wherein the first image processing scheme implements a first machine learning (ML) algorithm configured to detect plant objects or target objects on a portion of the first image;   detect a first target object in a first portion of the first image, the detection comprising identifying at least a first class associated with the first target object;   access the first portion of the first image for further processing;   apply a second image processing scheme on the first portion of the first image, wherein the second image processing scheme implements a second ML algorithm configured to classify a received image, the received image including the first portion of the first image;   identify at least a second class associated with the first portion of the first image;   compare results of the first class with the second class; and   perform a subsequent action based on the comparison.