Carsten Steger’s research while affiliated with RISC Software GmbH and other places

The unsupervised detection and localization of anomalies in natural images is an intriguing and challenging problem. Anomalies manifest themselves in very different ways and an ideal benchmark dataset for this task should contain representative examples for all of them. We find that existing datasets are biased towards local structural anomalies such as scratches, dents, or contaminations. In particular, they lack anomalies in the form of violations of logical constraints, e.g., permissible objects occurring in invalid locations. We contribute a new dataset based on industrial inspection scenarios that evenly covers both types of anomalies. We provide pixel-precise ground truth data for each anomalous region and define a generalized evaluation metric that addresses localization ambiguities that can arise for logical anomalies. Furthermore, we propose a novel algorithm that improves over the state of the art in the joint detection of structural and logical anomalies. It consists of a local and a global network branch. The first one inspects confined regions independent of their spatial locations in the input image and is primarily responsible for the detection of entirely new local structures. The second one learns a globally consistent representation of the training data through a bottleneck that enables the detection of violations of long-range dependencies, a key characteristic of many logical anomalies. We perform extensive evaluations on our new dataset to corroborate our claims.

We propose a novel multi-view camera model for line-scan cameras with telecentric lenses. The camera model supports an arbitrary number of cameras and assumes a linear relative motion with constant velocity between the cameras and the object. We distinguish two motion configurations. In the first configuration, all cameras move with independent motion vectors. In the second configuration, the cameras are mounted rigidly with respect to each other and therefore share a common motion vector. The camera model can model arbitrary lens distortions by supporting arbitrary positions of the line sensor with respect to the optical axis. We propose an algorithm to calibrate a multi-view telecentric line-scan camera setup. To facilitate a 3D reconstruction, we prove that an image pair acquired with two telecentric line-scan cameras can always be rectified to the epipolar standard configuration, in contrast to line-scan cameras with entocentric lenses, for which this is possible only under very restricted conditions. The rectification allows an arbitrary stereo algorithm to be used to calculate disparity images. We propose an efficient algorithm to compute 3D coordinates from these disparities. Experiments on real images show the validity of the proposed multi-view telecentric line-scan camera model.

We introduce the first comprehensive 3D dataset for the task of unsupervised anomaly detection and localization. It is inspired by real-world visual inspection scenarios in which a model has to detect various types of defects on manufactured products, even if it is trained only on anomaly-free data. There are defects that manifest themselves as anomalies in the geometric structure of an object. These cause significant deviations in a 3D representation of the data. We employed a high-resolution industrial 3D sensor to acquire depth scans of 10 different object categories. For all object categories, we present a training and validation set, each of which solely consists of scans of anomaly-free samples. The corresponding test sets contain samples showing various defects such as scratches, dents, holes, contaminations, or deformations. Precise ground-truth annotations are provided for every anomalous test sample. An initial benchmark of 3D anomaly detection methods on our dataset indicates a considerable room for improvement.

We present the shape model object tracker, which is accurate, robust, and real-time capable on a standard CPU. The tracker has a failure mode detection, is robust to nonlinear illumination changes, and can cope with occlusions. It uses subpixel-precise image edges to track roughly rigid objects with high accuracy and is virtually drift-free even for long sequences. Furthermore, it is inherently capable of object re-detection when tracking fails. To evaluate the accuracy, robustness, and efficiency of the tracker precisely, we present a challenging new tracking dataset with pixel-precise ground truth. The precise ground-truth labels are created automatically from the photo-realistic synthetic VIPER dataset. The tracker is thoroughly evaluated against the state of the art through a number of qualitative and quantitative experiments. It is able to perform on par with the current state-of-the-art deep-learning trackers, but is at least 45 times faster, even without using a GPU. The efficiency and low memory consumption of the tracker are validated in further experiments that are conducted on an embedded device.

The detection of anomalous structures in natural image data is of utmost importance for numerous tasks in the field of computer vision. The development of methods for unsupervised anomaly detection requires data on which to train and evaluate new approaches and ideas. We introduce the MVTec anomaly detection dataset containing 5354 high-resolution color images of different object and texture categories. It contains normal, i.e., defect-free images intended for training and images with anomalies intended for testing. The anomalies manifest themselves in the form of over 70 different types of defects such as scratches, dents, contaminations, and various structural changes. In addition, we provide pixel-precise ground truth annotations for all anomalies. We conduct a thorough evaluation of current state-of-the-art unsupervised anomaly detection methods based on deep architectures such as convolutional autoencoders, generative adversarial networks, and feature descriptors using pretrained convolutional neural networks, as well as classical computer vision methods. We highlight the advantages and disadvantages of multiple performance metrics as well as threshold estimation techniques. This benchmark indicates that methods that leverage descriptors of pretrained networks outperform all other approaches and deep-learning-based generative models show considerable room for improvement.

We propose a camera model for line-scan cameras with telecentric lenses. The camera model assumes a linear relative motion with constant velocity between the camera and the object. It allows to model lens distortions, while supporting arbitrary positions of the line sensor with respect to the optical axis. We comprehensively examine the degeneracies of the camera model and propose methods to handle them. Furthermore, we examine the relation of the proposed camera model to affine cameras. In addition, we propose an algorithm to calibrate telecentric line-scan cameras using a planar calibration object. We perform an extensive evaluation of the proposed camera model that establishes the validity and accuracy of the proposed model. We also show that even for lenses with very small lens distortions, the distortions are statistically highly significant. Therefore, they cannot be omitted in real-world applications.

We introduce a powerful student-teacher framework for the challenging problem of unsupervised anomaly detection and pixel-precise anomaly segmentation in high-resolution images. Student networks are trained to regress the output of a descriptive teacher network that was pretrained on a large dataset of patches from natural images. This circumvents the need for prior data annotation. Anomalies are detected when the outputs of the student networks differ from that of the teacher network. This happens when they fail to generalize outside the manifold of anomaly-free training data. The intrinsic uncertainty in the student networks is used as an additional scoring function that indicates anomalies. We compare our method to a large number of existing deep learning based methods for unsupervised anomaly detection. Our experiments demonstrate improvements over state-of-the-art methods on a number of real-world datasets, including the recently introduced MVTec Anomaly Detection dataset that was specifically designed to benchmark anomaly segmentation algorithms.

We introduce a powerful student-teacher framework for the challenging problem of unsupervised anomaly detection and pixel-precise anomaly segmentation in high-resolution images. Student networks are trained to regress the output of a descriptive teacher network that was pretrained on a large dataset of patches from natural images. This circumvents the need for prior data annotation. Anomalies are detected when the outputs of the student networks differ from that of the teacher network. This happens when they fail to generalize outside the manifold of anomaly-free training data. The intrinsic uncertainty in the student networks is used as an additional scoring function that indicates anomalies. We compare our method to a large number of existing deep learning based methods for unsupervised anomaly detection. Our experiments demonstrate improvements over state-of-the-art methods on a number of real-world datasets, including the recently introduced MVTec Anomaly Detection dataset that was specifically designed to benchmark anomaly segmentation algorithms.