Methods, systems, articles of manufacture, and apparatus to recalibrate confidences for image classification

Patent metadata

Field	Value
Publication number	US-12561400-B2
Application number	US-202418634433-A
Country	US
Kind code	B2
Filing date	Apr 12, 2024
Priority date	Nov 23, 2020
Publication date	Feb 24, 2026
Grant date	Feb 24, 2026

Abstract

Methods, systems, articles of manufacture, and apparatus to recalibrate confidences for image classification are disclosed. An example apparatus to classify an image includes an image crop detector to detect a first image crop from the image, the first image crop corresponding to a first object, a grouping controller to select a second image crop corresponding to a second object at a location of the first object, a prediction generator to, in response to executing a trained model, determine a label corresponding to the first object and a confidence level associated with the label, and a confidence recalibrator to recalibrate the confidence level based on a probability of the first object having a first attribute based on the second object having a second attribute, the confidence level recalibrated to increase an accuracy of the image classification.

First claim

What is claimed is:

1 . An apparatus comprising:
memory; machine readable instructions; and at least one processor circuit to be programmed by the machine readable instructions to:
access labeled image data associated with images of objects on shelves;
generate a co-occurrence distribution graph based on the labeled image data, the co-occurrence distribution graph representing a probability distribution of different pairs of characteristics of the objects occurring on a same one of the shelves; and
classify a characteristic of a first object in an image crop based on the co-occurrence distribution graph; and
generate a report for a user including the characteristic of the first object.

2 . The apparatus of claim 1 , wherein one or more of the at least one processor circuit is to:
generate a co-occurrence cardinality graph based on the labeled image data, the co-occurrence cardinality graph including numbers of instances of the different pairs of characteristics of the objects occurring on the same one of the shelves; and generate the co-occurrence distribution graph based on the co-occurrence cardinality graph.

3 . The apparatus of claim 2 , wherein one or more of the at least one processor circuit is to generate the co-occurrence distribution graph by normalizing the co-occurrence cardinality graph.

4 . The apparatus of claim 3 , wherein the co-occurrence cardinality graph incudes an N-by-N matrix, where N represents the number of unique characteristics represented in the labeled image data.

5 . The apparatus of claim 4 , wherein one or more of the at least one processor circuit is to determine normalized values in a first column of the co-occurrence distribution graph by determining a sum of values from a corresponding first column of the co-occurrence graph and dividing each of the values in the first column of the co-occurrence cardinality graph by the sum.

6 . The apparatus of claim 2 , wherein the labeled image data includes locations of the shelves corresponding to objects in the images, and wherein one or more of the at least one processor circuit is to:
group the images into a first group of images with objects on a first one of the shelves and a second group of images with objects on a second one of the shelves based on the labeled image data; and generate the co-occurrence cardinality graph based on the groupings of the first group and the second group.

7 . The apparatus of claim 1 , wherein the characteristic is a volume, and wherein the co-occurrence distribution graph is a volume co-occurrence distribution graph.

8 . The apparatus of claim 1 , wherein the first object is a product, and wherein the image crop is from an image of a shelf in a store having the product.

9 . The apparatus of claim 1 , wherein one or more of the at least one processor circuit is to classify the characteristic of the first object in the image crop by:
executing a machine learning model to generate a plurality of characteristic predictions for the first object and a plurality of confidence levels associated with respective ones of the plurality of characteristic predictions.

10 . The apparatus of claim 9 , wherein one or more of the at least one processor circuit is to update the confidence levels based on the co-occurrence distribution graph.

11 . A non-transitory computer readable medium comprising instructions which, when executed, cause at least one processor circuit to:
access labeled image data associated with images of objects on shelves; generate a co-occurrence distribution graph based on the labeled image data, the co-occurrence distribution graph representing a probability distribution of different pairs of characteristics of the objects occurring on a same one of the shelves; and classify a characteristic of a first object in an image crop based on the co-occurrence distribution graph; and generate a report for a user including the characteristic of the first object.

12 . The non-transitory computer readable medium of claim 11 , wherein the instructions cause the at least one processor circuit to:
generate a co-occurrence cardinality graph based on the labeled image data, the co-occurrence cardinality graph including numbers of instances of the different pairs of characteristics of the objects occurring on the same one of the shelves; and generate the co-occurrence distribution graph based on the co-occurrence cardinality graph.

13 . The non-transitory computer readable medium of claim 12 , wherein the instructions cause the at least one processor circuit to generate the co-occurrence distribution graph by normalizing the co-occurrence cardinality graph.

14 . The non-transitory computer readable medium of claim 13 , wherein the co-occurrence cardinality graph incudes an N-by-N matrix, where N represents the number of unique characteristics represented in the labeled image data.

15 . The non-transitory computer readable medium of claim 14 , wherein the instructions cause the at least one processor circuit to determine normalized values in a first column of the co-occurrence distribution graph by determining a sum of values from a corresponding first column of the co-occurrence graph and dividing each of the values in the first column of the co-occurrence cardinality graph by the sum.

16 . The non-transitory computer readable medium of claim 12 , wherein the labeled image data includes locations of the shelves corresponding to objects in the images, and wherein the instructions cause the at least one processor circuit to:
group the images into a first group of images with objects on a first one of the shelves and a second group of images with objects on a second one of the shelves based on the labeled image data; and generate the co-occurrence cardinality graph based on the groupings of the first group and the second group.

17 . The non-transitory computer readable medium of claim 11 , wherein the characteristic is a volume, and wherein the co-occurrence distribution graph is a volume co-occurrence distribution graph.

18 . The non-transitory computer readable medium of claim 11 , wherein the first object is a product, and wherein the image crop is from an image of a shelf in a store having the product.

19 . The non-transitory computer readable medium of claim 11 , wherein the instructions cause the at least one processor circuit to classify the characteristic of the first object in the image crop by:
executing a machine learning model to generate a plurality of characteristic predictions for the first object and a plurality of confidence levels associated with respective ones of the plurality of characteristic predictions.

20 . The non-transitory computer readable medium of claim 19 , wherein the instructions cause the at least one processor circuit to update the confidence levels based on the co-occurrence distribution graph.

Assignees

• Nielsen Consumer Llc

Inventors

• Almazan Emilio
• Aller Beascoechea Aitor
• Tovar Velasco Javier

Classifications

• G06N3/08

  Learning methods · CPC title

• G06N3/04

  Architecture, e.g. interconnection topology · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06F18/2415

  based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title