Systems and methods for enhancing real-time image recognition

Patent metadata

Field	Value
Publication number	US-12340578-B2
Application number	US-202318452057-A
Country	US
Kind code	B2
Filing date	Aug 18, 2023
Priority date	May 29, 2019
Publication date	Jun 24, 2025
Grant date	Jun 24, 2025

Abstract

Disclosed embodiments provide systems, methods, and computer-readable storage media for enhancing a vehicle identification with preprocessing. The system may comprise memory and processor devices to execute instructions for receiving an image depicting a vehicle. The image may be analyzed and first predicted identity and first confidence value may be determined. The first confidence value may be compared to a predetermined threshold. The processors may further select a processing technique for modifying the image and further analyze the modified image determining a second predicted identity of the vehicle. And a second confidence value may be determined. And the system may further compare the second confidence value to the predetermined threshold to select the first or second predicted identity for transmission to a user.

First claim

What is claimed is:

1. A system for identifying objects in images, the system comprising:
at least one memory device storing instructions;
an augmentation tool configured to select at least one processing technique for modifying images; and
at least one processor configured to execute instructions to:
receive an image of an object;
analyze the image of an object to determine a first predicted identity depicted in the image by applying, to the object, an object identification model comprising at least one of a neural network model, a regression model, a statistical model, a parametric model, a non-parametric model, or a semi-parametric model;
determine a first confidence value distribution associated with the first predicted identity;
modify the image using the augmentation tool and at least one processing technique;
analyze the modified image to determine a second predicted identity depicted in the modified image by applying, to the modified image, the object identification model;
determine a second confidence value associated with the second predicted identity;
perform a comparison of at least one of:
a value in the second confidence value distribution to a predetermined threshold value, or
the first confidence value distribution to the second confidence value distribution;

generate an output for transmission to a user based on the comparison; and
transmit the output to the user.

2. The system of claim 1 , wherein the object identification model is configured to recognize a sub-group of classes containing objects.

3. The system of claim 1 , wherein the object identification model includes a convolutional neural network for detecting objects and determining attributes in an image based on extracted features from the image.

4. The system of claim 3 , wherein the object identification model comprises at least one of a You Only Look Once (YOLO) object detection framework, a Single-Shot Detector (SSD) object detection framework, an Inception object detection framework, a Visual Geometry Group (VGG) object detection framework, or a Residual Neural Network (ResNet) object detection framework.

5. The system of claim 1 , wherein the object identification model is configured to output class identification conclusions based on assigned confidence scores.

6. The system of claim 1 , wherein the object identification model is configured to sort elements of a dataset using one or more classifiers to determine a probability of an outcome.

7. The system of claim 1 , wherein the object includes at least one of a vehicle, a tree, a cat, a dog, or a human.

8. The system of claim 1 , wherein modifying the image comprises at least one of adjusting a color intensity of at least one pixel of the image, adjusting a hue of at least one pixel of the image, adjusting a brightness of at least one pixel of the image, adjusting a contrast of the image, or adjusting an orientation of the image.

9. The system of claim 1 , wherein the augmentation tool includes a machine learning model configured to determine an attribute for modification of an object based on historical modification data.

10. The system of claim 1 , wherein the object identification model is usable by an identification model application installed on a user device; and wherein the user device is unable to modify the object identification model.

11. A user device comprising:
at least one memory device storing instructions;
an image sensor;
an augmentation tool configured to select at least one processing technique for modifying images; and
at least one processor configured to execute instructions to:
receive an image of an object;
analyze the image of an object to determine a first predicted identity depicted in the image by applying, to the object, an object identification model comprising at least one of a neural network model, a regression model, a statistical model, a parametric model, a non-parametric model, or a semi-parametric model;
determine a first confidence value distribution associated with the first predicted identity;
modify the image using the augmentation tool and at least one processing technique;
analyze the modified image to determine a second predicted identity depicted in the modified image by applying, to the modified image, the object identification model;
determine a second confidence value associated with the second predicted identity;
perform a comparison of at least one of:
a value in the second confidence value distribution to a predetermined threshold value, or
the first confidence value distribution to the second confidence value distribution;

generate an output for transmission to a user based on the comparison; and
transmit the output to the user.

12. The user device of claim 11 , wherein the object identification model is configured to recognize a sub-group of classes containing objects.

13. The user device of claim 11 , wherein the object identification model includes a convolutional neural network for detecting objects and determining attributes in an image based on extracted features from the image.

14. The user device of claim 13 , wherein the object identification model comprises at least one of: a You Only Look Once (YOLO) object detection framework, a Single-Shot Detector (SSD) object detection framework, an Inception object detection framework, a Visual Geometry Group (VGG) object detection framework, and a Residual Neural Network (ResNet) object detection framework.

15. The user device of claim 11 , wherein the object identification model is configured to output class identification conclusions based on assigned confidence scores.

16. The user device of claim 15 , wherein the object identification model is configured to sort elements of a dataset using one or more classifiers to determine a probability of an outcome.

17. The user device of claim 11 , wherein the object comprises at least one of a vehicle, a tree, a cat, a dog, or a human.

18. The user device of claim 11 , wherein modifying the image comprises at least one of adjusting a color intensity of at least one pixel of the image, adjusting a hue of at least one pixel of the image, adjusting a brightness of at least one pixel of the image, adjusting a contrast of the image, or adjusting an orientation of the image.

19. The user device of claim 11 , wherein the object identification model is usable by an identification model application installed on a user device; and wherein the user device is unable to modify the object identification model.

20. The user device of claim 11 , wherein the augmentation tool is configured to dynamically select the at least one processing technique based on an augmentation optimizer model.

Assignees

• Capital One Services Llc

Inventors

• Price Micah
• Ho Chi-San
• Charania Aamer
• Vasisht Sunil

Classifications

• G06T11/10

  Texturing; Colouring; Generation of textures or colours (retouching, inpainting or scratch removal G06T5/77) · CPC title

• G06T5/90

  Dynamic range modification of images or parts thereof · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06V2201/08

  Detecting or categorising vehicles · CPC title