# US 12,001,917 B2 · Hub-and-spoke classification system and methods

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,001,917 B2
- Document ID: US12001917B2
- Status: Granted
- Kind: B2
- Application number: 16/818,411
- Filed: 2020-03-13
- Published: 2020-09-10
- Granted: 2024-06-04
- Earliest claimed: 2017-06-05
- Assignees: Umajin Inc.
- Applicants: Umajin Inc.
- Inventors: David Brebner
- Web page: https://patentgrep.com/patent/US12001917B2

## CPC

- G06N 20/20: Ensemble learning
- G06F 16/55: Clustering; Classification
- G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F 18/22: Matching criteria, e.g. proximity measures
- G06F 18/24: Classification techniques
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/0475: Generative networks
- G06N 3/0495: Quantised networks; Sparse networks; Compressed networks
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 3/098: Distributed learning, e.g. federated learning
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 5/048: Fuzzy inferencing
- G06N 20/00: Machine learning
- G06V 10/82: using neural networks
- G06V 20/20: in augmented reality scenes
- G06V 30/19173: {Classification techniques}
- G06V 40/28: {Recognition of hand or arm movements, e.g. recognition of deaf sign language (static hand signs G06V40/113)}
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/045: Combinations of networks
- G06N 3/08: Learning methods
- G06N 3/126: Evolutionary algorithms, e.g. genetic algorithms or genetic programming
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 7/023: {Learning or tuning the parameters of a fuzzy system}
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

According to some embodiments of the present disclosure, the disclosure relates to an application system and server kit that create and serve digital twin-enabled applications. This disclosure also relates to a hub-and-spoke classification system. This disclosure also relates to a location-based services framework that leverages a generative content process to improve location prediction. This disclosure also relates to virtual reality and augmented reality applications, as well as digital agents that support various types of applications.

## Claims

### Claim 1

A method for training a classification device, comprising:
receiving, by one or more processors of a hub device, a plurality of images of an object being monitored by the classification device;
extracting, by the one or more processors, a set of image subsections from the plurality of images;
determining, by the one or more processors, a set of potential templates from the set of image subsections;
for each potential template of the set of potential templates:
determining, by the one or more processors, a set of weighted match scores for the potential template, wherein each of the weighted match scores corresponds to a respective classification and indicates a degree of confidence in the respective classification;
determining, by the one or more processors, a set of similarity scores for the potential template, wherein each of the similarity scores corresponds to another respective image subsection from the set of image subsections and indicates a degree of similarity between the potential template and the other respective image subsection;
in response to determining one of the weighted match scores exceeds a match score threshold and at least one of the similarity scores exceeds a first similarity threshold:
assigning, by the one or more processors, the respective classification corresponding to the weighted match score that exceeded the match score threshold to the potential template; and
including, by the one or more processors, the potential template in a set of classifiable templates corresponding to the classification device;
in response to determining one of the weighted match scores exceeds a match score threshold and a plurality of the similarity scores are less than the first similarity threshold and greater than a second similarity threshold:
assigning, by the one or more processors, the respective classification corresponding to the weighted match score that exceeded the match score threshold to the potential template; and
including, by the one or more processors, the potential template in a set of alternative templates corresponding to the classification device; and
providing the set of alternative templates and the set of classifiable templates to the classification device.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
in response to determining none of the similarity scores are greater than a second similarity threshold:
generating a set of synthesized templates based on the set of classifiable templates, wherein each synthesized template includes a synthesized image containing two or more character fragments from two or more respective classifiable templates and a synthesized classification that indicates a character represented by the synthesized template; and
for each synthesized template in the set of synthesized templates:
determining a synthesized template similarity score between the synthesized template and the potential template;
in response to determining that the synthesized template similarity score of the synthesized template exceeds a synthesized template threshold:
assigning the synthesized classification to the potential template; and
including the potential template in a set of character fragment templates; and
providing the set of character fragment templates to the image classification device.

### Claim 3 (depends on 2)

The method of claim 2, wherein each synthesized template of the set of synthesized templates represents a respective transition between two characters.

### Claim 4 (depends on 3)

The method of claim 3, wherein each respective transition between two characters is a transition between the two characters on a wheel counter meter.

### Claim 5 (depends on 1)

The method of claim 1, wherein determining the set of weighted match scores for each potential template includes performing optical character recognition on the image subsection corresponding to the potential template, wherein the optical character recognition provides a set of potential matches and, for each potential match, a weighted match score of the potential match with respect to the image subsection.

### Claim 6 (depends on 1)

The method of claim 1, wherein determining the set of similarity scores for the potential template includes, for each of the other respective image subsections, calculating a ratio of matching pixels between the image subsection corresponding to the potential template and the other respective image subsections.

### Claim 7 (depends on 1)

The method of claim 1, wherein the plurality of images are received in response to commanding the classification device to capture the plurality of images.

### Claim 8 (depends on 1)

The method of claim 1, wherein extracting the set of image subsections includes, for each image in the plurality of images:
extracting a predefined region of the image corresponding to the object being monitored; and
extracting one or more image subsections from the predefined region of the image.

### Claim 9 (depends on 8)

The method of claim 8, wherein the predefined region is defined by a bounding box defined by a human using a configuration device.

### Claim 10 (depends on 8)

The method of claim 8, wherein the image subsections extracted from the predefined region are extracted from predefined subsections of the predefined region.

### Claim 11 (depends on 1)

The method of claim 1, wherein determining the set of potential templates based on the set of image subsections includes identifying any subsets of matching image subsections from the set of image subsections, wherein only one image subsection from each subset of matching images is included in the set of potential templates.

### Claim 12 (depends on 11)

The method of claim 11, wherein each potential template is initially unclassified.

### Claim 13 (depends on 1)

The method of claim 1, wherein the object being monitored by the image classification device is a meter.

### Claim 14 (depends on 13)

The method of claim 13, wherein the meter is a wheel counter meter.

### Claim 15 (depends on 14)

The method of claim 14, wherein the meter is an LED meter.

### Claim 16 (depends on 1)

The method of claim 1, wherein the hub device trains multiple classification devices.

### Claim 17 (depends on 1)

The method of claim 1, wherein the hub device is connected to a continuous power supply and the classification device is powered by a battery.