IP Library Granted Patent US 12694508
Granted Patent B2
US 12694508 · App. 18/511,337 · Granted Jul 28, 2026

Anomaly detection method, anomaly detection device, and recording medium

Inventors: Denis Gudovskiy (San Ramon, CA); Shun Ishizaka (Tokyo, JP); Kazuki Kozuka (Osaka, JP)
Assignee: PANASONIC INTELLECTUAL PROPERTY CORPORATION OF AMERICA
G06T7/001G06N3/0464G06V10/454G06T2207/20084
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12694508
App. No.
18/511,337
Granted
Jul 28, 2026
Kind
B2
Abstract

An anomaly detection method by which a computer performs anomaly detection includes: obtaining first feature data outputted through N (N is an integer not less than 1) convolutional layers of a convolutional neural network configured as an encoder when an image is inputted to the convolutional neural network; obtaining second feature data outputted through M (M is an integer not less than 1, and M≠N) convolutional layers of the convolutional neural network and different in size from the first feature data; and performing anomaly detection on the image by using features indicated by the first feature data and the second feature data that are different in size.

Claims (22)

1 . An anomaly detection method by which a computer performs anomaly detection, the method comprising:

obtaining first feature data outputted through N convolutional layers of a convolutional neural network configured as an encoder when an image is inputted to the convolutional neural network, N being an integer not less than 1;

obtaining second feature data outputted through M convolutional layers of the convolutional neural network and different in size from the first feature data, M being an integer not less than 1 and not equal to N; and

performing anomaly detection on the image by using features indicated by the first feature data and the second feature data that are different in size, wherein

at least part of the M convolutional layers is shared with at least part of the N convolutional layers.

2 . The anomaly detection method according to claim 1 , wherein

the performing of the anomaly detection includes:

obtaining first output data outputted from a first decoder including a neural network, by inputting the first feature data to the first decoder;

obtaining second output data outputted from a second decoder including a neural network, by inputting the second feature data to the second decoder; and

generating and outputting an output image indicating a location in which an anomaly in the image is shown, by aggregating the first output data and the second output data.

3 . The anomaly detection method according to claim 2 , wherein

the first decoder and the second decoder are trained to output first output data and second output data that satisfy a predetermined condition, respectively, in response to input of first feature data and second feature data obtained with respect to a normal image in which no anomaly is shown, and

the predetermined condition is that an output image indicating no anomaly is generated by aggregating the first output data and the second output data.

4 . A non-transitory computer-readable recording medium having recorded thereon a program for causing a computer to execute the anomaly detection method according to claim 1 .

5 . An anomaly detection device in which a computer performs anomaly detection, the device comprising:

a processor; and

memory connected to the processor, wherein

using the memory, the processor:

obtains first feature data outputted through N convolutional layers of a convolutional neural network configured as an encoder when an image is inputted to the convolutional neural network, N being an integer not less than 1;

obtains second feature data outputted through M convolutional layers of the convolutional neural network and different in size from the first feature data, M being an integer not less than 1 and not equal to N; and

performs anomaly detection on the image by using features indicated by the first feature data and the second feature data that are different in size, wherein

at least part of the M convolutional layers is shared with at least part of the N convolutional layers.