Integrating machine learning classification models and machine learning anomaly models
Inventors
Villemaire, André • Savary, Simon • GARDNER, MARC-ANDRÉ • Bloch, Olivier
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12518202-B1
Publication Date
2026-01-06
Expiration Date
Abstract
In example embodiments, a hybrid classification/anomaly machine learning architecture is provided that combines a classification model and an anomaly model to perform an engineering task. The classification model and the anomaly model may be used in parallel, and their inference results compared, with their consistency used to improve confidence, and their inconsistency used to detect when additional training or other improvement is required and to capture data useful in such additional training/improvement.
Core Innovation
The invention inspects infrastructure for damage by using a hybrid classification/anomaly machine learning architecture. Input data that includes representations of infrastructure is accessed, and the input data is submitted in parallel to a classification model and an anomaly model executed on one or more computing devices.
The classification model produces a first inference result indicating whether a class of damage is detected in the representations of infrastructure. The anomaly model produces a second inference result indicating whether something other than undamaged condition is detected in the representations of infrastructure. An application compares the first and second inference results for consistency based on whether the damage-class result matches the anomaly result.
When consistency between the first and second inference results yields consistent positive results, the application reports the class of damage to the infrastructure in a user interface. In response to consistency or inconsistency, the invention updates training data and dataset components, including adding input data to a feature dataset and/or an anomaly dataset, and optionally using subject matter expert assessments to inject ground truth.
Claims Coverage
The document includes two independent claims, a method and a system. Across these claims, there are 3 inventive features focused on parallel classification/anomaly inference, consistency-based comparison, and response actions that include reporting and dataset updates.
Parallel hybrid classification and anomaly inference for infrastructure representations
Submitting input data in parallel to a classification model and an anomaly model of a classification/anomaly machine learning architecture executing on one or more computing devices, wherein the input data includes representations of infrastructure including at least one of bridges, buildings, roads, railways, electrical and communications networks, or equipment; producing a first inference result indicating whether a class of damage is detected and producing a second inference result indicating whether something other than undamaged condition is detected.
Consistency comparison of classification and anomaly inference results
Comparing the first inference result and the second inference result for consistency in which the first inference result and the second inference result are considered to have consistency when the first inference result indicates a class of damage is detected and the second inference result indicates something other than undamaged condition is detected or when the first inference result indicates no class of damage is detected and the second inference result indicates undamaged condition is detected.
Consistency-driven reporting or dataset update
In response to consistency between the first inference result and the second inference result, reporting the class of damage to the infrastructure in a user interface and/or performing at least one of reporting a detected feature or adding the input data to a feature dataset used to train the classification model and/or adding the input data to an anomaly dataset used to train the anomaly model.
Both independent claims define a hybrid architecture that runs a classification model and an anomaly model in parallel, compares their inference results for consistency using a damage/undamaged condition matching rule, and then responds to consistency by reporting and/or updating feature and/or anomaly datasets for training.
Stated Advantages
Improves confidence in reporting by requiring consistency between the classification inference result and the anomaly inference result.
Documented Applications
Inspecting infrastructure for damage in bridges, buildings, roads, railways, electrical and communications networks, and equipment, including reporting detected damage classes in a user interface.
Interested in licensing this patent?