Please note that the ITEA Office will be closed from 25 December 2024 to 1 January 2025 inclusive.
Discrepancy Scaling for Unsupervised Anomaly Detection and Localization
- Project
- 20219 IML4E
- Type
- New service
- Description
Discrepancy Scaling is a fast, accurate and lightweight deep learning-based method for unsupervised anomaly detection (AD) and anomaly localization (AL) in images.
- Contact
- Juha Mylläri
- juha.myllari@helsinki.fi
- Research area(s)
- Computer vision, deep learning, unsupervised learning.
- Technical features
Discrepancy Scaling uses convolutional neural networks to identify anomalous images and segment them into anomalous and normal regions. It can be applied to both natural images (photographs) and artificial images such as spectrograms. The method is trained using normal images only.
- Integration constraints
Discrepancy Scaling is implemented in Python 3 using the PyTorch deep learning library.
- Targeted customer(s)
The method is particularly well suited for quality management in manufacturing but also has applications in construction, maintenance etc.
- Conditions for reuse
The code for Discrepancy Scaling is released under the GPL-3.0 license.
- Confidentiality
- Public
- Publication date
- 20-09-2024
- Involved partners
- University of Helsinki (FIN)
