System for learning new visual inspection tasks using a few-shot meta-learning method

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published April 7th, 2026

Detected April 8th, 2026

Summary

USPTO granted patent US12597232B2 to Hitachi, Ltd. covering a system and method for few-shot meta-learning to enable visual inspection models to adapt to new domains with minimal labeled training data. The patent describes processing labeled and unlabeled images through multiple backbone snapshots, extracting features, and generating representative models through clustering and transformation techniques. This machine learning patent covers neural network architectures applicable to automated quality control and inspection systems.

View original document View source feed page

What changed

USPTO granted patent US12597232B2 to Hitachi, Ltd. for a few-shot meta-learning system enabling visual inspection models to learn new tasks from minimal labeled data. The invention processes labeled and unlabeled images through backbone snapshots trained across multiple domain tasks, extracts features using neural networks, and generates representative inspection models through feature clustering and transformation. The patent covers G06N (neural network) and G06V (computer vision) classification areas.

The patent grants Hitachi exclusive rights to this AI approach for visual inspection, which may affect technology companies developing machine learning-based quality control, manufacturing inspection, or defect detection systems. Companies in manufacturing, electronics, automotive, and other industries utilizing automated visual inspection should monitor for potential licensing requirements or design-around considerations.

What to do next

Monitor for updates

Source document (simplified)

← USPTO Patent Grants

System for learning new visual inspection tasks using a few-shot meta-learning method

Grant US12597232B2 Kind: B2 Apr 07, 2026

Assignee

HITACHI, LTD.

Inventors

Lasitha Sandaruwan Vidyaratne, Xian Yeow Lee, Mahbubul Alam, Ahmed Farahat, Dipanjan Ghosh, Maria Teresa Gonzalez Diaz, Chetan Gupta

Abstract

Systems and methods described herein which can involve for a first input of a plurality of labeled images of a new domain task, processing the first plurality of labeled images through a plurality of backbone snapshots, each of the backbone snapshots representative of a model trained across a plurality of other domain tasks, each of the plurality of backbone snapshots configured to output a first plurality of features responsive to the input; processing a second input of second plurality of unlabeled images through the plurality of backbone snapshots to output a second plurality of features responsive to the second input; and generating a representative model for the new domain task from the clustering and transformation of the first plurality of features and as associated from the clustered and transformed second plurality of features.

CPC Classifications

G06N 3/02 G06N 3/08 G06N 3/042 G06N 3/045 G06N 3/047 G06N 3/082 G06N 3/092 G06N 3/0454 G06N 3/0475 G06N 3/0464 G06N 3/088 G06N 3/0499 G06N 3/096 G06N 3/0985 G06N 20/00 G06N 20/10 G06N 20/20 G06V 10/82 G06V 10/761 G06V 10/762 G06V 10/7715