Fast and Efficient Image Novelty Detection Based on Mean-Shifts

Cite This

Files in this item

Checksum: MD5:074fe53772fbdc803b48c46d2370efd3

HERMANN, Matthias, Georg UMLAUF, Bastian GOLDLÜCKE, Matthias O. FRANZ, 2022. Fast and Efficient Image Novelty Detection Based on Mean-Shifts. In: Sensors. MDPI. 22(19), 7674. eISSN 1424-8220. Available under: doi: 10.3390/s22197674

@article{Hermann2022-10-10Effic-58928, title={Fast and Efficient Image Novelty Detection Based on Mean-Shifts}, year={2022}, doi={10.3390/s22197674}, number={19}, volume={22}, journal={Sensors}, author={Hermann, Matthias and Umlauf, Georg and Goldlücke, Bastian and Franz, Matthias O.}, note={Article Number: 7674} }

Hermann, Matthias 2022-10-10 Umlauf, Georg Goldlücke, Bastian Attribution 4.0 International Franz, Matthias O. 2022-10-27T09:31:09Z Franz, Matthias O. Hermann, Matthias Image novelty detection is a repeating task in computer vision and describes the detection of anomalous images based on a training dataset consisting solely of normal reference data. It has been found that, in particular, neural networks are well-suited for the task. Our approach first transforms the training and test images into ensembles of patches, which enables the assessment of mean-shifts between normal data and outliers. As mean-shifts are only detectable when the outlier ensemble and inlier distribution are spatially separate from each other, a rich feature space, such as a pre-trained neural network, needs to be chosen to represent the extracted patches. For mean-shift estimation, the Hotelling T<sup>2</sup> test is used. The size of the patches turned out to be a crucial hyperparameter that needs additional domain knowledge about the spatial size of the expected anomalies (local vs. global). This also affects model selection and the chosen feature space, as commonly used Convolutional Neural Networks or Vision Image Transformers have very different receptive field sizes. To showcase the state-of-the-art capabilities of our approach, we compare results with classical and deep learning methods on the popular dataset CIFAR-10, and demonstrate its real-world applicability in a large-scale industrial inspection scenario using the MVTec dataset. Because of the inexpensive design, our method can be implemented by a single additional 2D-convolution and pooling layer and allows particularly fast prediction times while being very data-efficient. eng Fast and Efficient Image Novelty Detection Based on Mean-Shifts Goldlücke, Bastian Umlauf, Georg 2022-10-27T09:31:09Z

Downloads since Oct 27, 2022 (Information about access statistics)

Hermann_2-1q5sij66uivsn1.pdf 14

This item appears in the following Collection(s)

Attribution 4.0 International Except where otherwise noted, this item's license is described as Attribution 4.0 International

Search KOPS


Browse

My Account