- Media recognition
Image anomaly detection technology using difference between an inspection image and reference images
Image-based anomaly detection method that requires no learning and can be applied immediately.
- Detects anomalies based on differences an between inspection image and reference images in the deep feature space.
- By subtracting the difference between reference images from an anomaly score map obtained using conventional methods, minimize false positives that often occur where differences are shown in edges or structures even if nothing is wrong.
Applications
- Saves labor in visual inspections of infrastructures.
Benchmarks, strengths, and track record
- Can be tested immediately with no need for learning if reference images are available.
Inquiries
Contact the Toshiba Corporate Research & Development Center
Please include the title “Toshiba AI Technology Catalog: Image anomaly detection technology using difference detection” or the URL in the inquiry text.
Please note that because this technology is currently the subject of R&D activities, immediate responses to inquiries may not be possible.
References:
- Kawamura, “Unsupervised Anomaly Localization using Locally Adaptive Query-Dependent Scores”, ICIAP2021, May, 2022
- Toshiba Develops Infrastructure Inspection AI for Detecting Anomalies from Only a Few Reference Images with Unprecedented Accuracy - Ease of incorporation into real-world inspections and accurate detection of infrequent, untrained anomalies helps reduce workloads, expedite discoveries, and ensure stable long-term operation of infrastructure - R&D Center; Toshiba (global.toshiba)
- Kawamura, “Unsupervised Anomaly Localization using Locally Adaptive Query-Dependent Scores”, ICIAP2021, May, 2022.
