Horst Bischof

Ein kompakter Blick auf Chancen, Risiken und Verantwortung der KI: Der Sammelband zeigt, wie die Steiermark Innovation, Ethik und europäische Regulierung vereint, um die digitale Zukunft aktiv zu gestalten. Gesamtes E-Book (kostenloser PDF-Download): Die einzelnen Kapitel finden Sie im unteren Reiter. Gedruckte Variante (Softcover - auf wenige Exemplare limitiert):

We propose a novel approach to video anomaly detection: we treat feature vectors extracted from videos as realizations of a random variable with a fixed distribution and model this distribution with a neural network. This lets us estimate the likelihood of test videos and detect video anomalies by thresholding the likelihood estimates. We train our video anomaly detector using a modification of denoising score matching, a method that injects training data with noise to facilitate modeling its distribution. To eliminate hyperparameter selection, we model the distribution of noisy video features across a range of noise levels and introduce a regularizer that tends to align the models for different levels of noise. At test time, we combine anomaly indications at multiple noise scales with a Gaussian mixture model. Running our video anomaly detector induces minimal delays as inference requires merely extracting the features and forward-propagating them through a shallow neural network and a Gaussian mixture model. Our experiments on five popular video anomaly detection benchmarks demonstrate state-of-the-art performance, both in the object-centric and in the frame-centric setup.

Bronchiolitis Obliterans Syndrome (BOS), a fibrotic airway disease following lung transplantation, conventionally relies on pulmonary function tests (PFTs) for diagnosis due to limitations of CT images. Thus far, deep neural networks (DNNs) have not been used for BOS detection. We optimized a DNN for detection of BOS solely using CT scans by integrating an innovative co-training method for enhanced performance in low-data scenarios. The novel auxiliary task is to predict the temporal precedence of CT scans of BOS patients. We tested our method using CT scans at various stages of inspiration from 75 post-transplant patients, including 26 with BOS. The method achieved a ROC-AUC of 0.90 (95% CI: 0.840-0.953) in distinguishing BOS from non-BOS CT scans. Performance correlated with disease progression, reaching 0.88 ROC-AUC for stage I, 0.91 for stage II, and an outstanding 0.94 for stage III BOS. Importantly, performance parity existed between standard and high-resolution scans. Particularly noteworthy is the DNN’s ability to predict BOS in at-risk patients (FEV1 between 80% and 90% of best FEV1) with a robust 0.87 ROC-AUC (CI: 0.735-0.974). Using techniques for visually interpreting the results of deep neural networks, we reveal that our method is especially sensitive to hyperlucent areas compatible with air-trapping or bronchiectasis. Our approach shows the potential to improve BOS diagnosis, enabling early detection and management. Detecting BOS from low-resolution scans reduces radiation exposure and using scans at any stage of respiration makes our method more accessible. Additionally, we demonstrate that techniques that limit overfitting are essential to unlocking the power of DNNs in scenarios with scarce training data. Our method may enable clinicians to use DNNs in studies where only a modest number of patients is available.

Dieser Artikel ist ein Plädoyer für die universitäre Forschung und insbesondere für forschungsstarke Universitäten. Der Autor legt dar, dass die Forschungsstärke ein ganz wesentliches Asset ist, und gerade im regionalen Eco-System eine ganz entscheidender Faktor ist. Das Konzept der Universität und der universitären Forschung ist zwar schon alt, Bologna und Oxford wurden 1088 bzw. 1096 gegründet, und dies sind nicht die allerersten Universitäten (um 800 gibt es bereits erste Vorläufer), aber das System funktioniert noch immer. Gerade die universitäre Forschung ist einem permanenten Prozess des sich Neuerfindens und der Erneuerung ausgesetzt und schafft es dadurch, immer am Puls der Zeit zu sein. In diesem Beitrag werden drei Aspekte ganz besonders beleuchtet: 1. Was ist das Besondere an der universitären Forschung? 2. Was sind die wesentlichen (modernen) Ingredienzien und welche Auswirkungen auf das lokale Eco-System haben diese? 3. Was kann eine Universitätsleitung dazu beitragen (und was nicht)? (DIPF/Orig.)

The sensing process of large-scale LiDAR point clouds inevitably causes large blind spots, i.e. regions not visible to the sensor. We demonstrate how these inherent sampling properties can be effectively utilized for self-supervised representation learning by designing a highly effective pre-training framework that considerably reduces the need for tedious 3D annotations to train state-of-the-art object detectors. Our Masked AutoEncoder for LiDAR point clouds (MAELi) intuitively leverages the sparsity of LiDAR point clouds in both the encoder and decoder during reconstruction. This results in more expressive and useful initialization, which can be directly applied to downstream perception tasks, such as 3D object detection or semantic segmentation for autonomous driving. In a novel reconstruction approach, MAELi distinguishes between empty and occluded space and employs a new masking strategy that targets the LiDAR's inherent spherical projection. Thereby, without any ground truth whatsoever and trained on single frames only, MAELi obtains an understanding of the underlying 3D scene geometry and semantics. To demonstrate the potential of MAELi, we pre-train backbones in an end-to-end manner and show the effectiveness of our unsupervised pre-trained weights on the tasks of 3D object detection and semantic segmentation.

Widely-used LiDAR-based 3D object detectors often neglect fundamental geometric information readily available from the object proposals in their confidence estimation. This is mostly due to architectural design choices, which were often adopted from the 2D image domain, where geometric context is rarely available. In 3D, however, considering the object properties and its surroundings in a holistic way is important to distinguish between true and false positive detections, e.g. occluded pedestrians in a group. To address this, we present GACE, an intuitive and highly efficient method to improve the confidence estimation of a given black-box 3D object detector. We aggregate geometric cues of detections and their spatial relationships, which enables us to properly assess their plausibility and consequently, improve the confidence estimation. This leads to consistent performance gains over a variety of state-of-the-art detectors. Across all evaluated detectors, GACE proves to be especially beneficial for the vulnerable road user classes, i.e. pedestrians and cyclists.

Test-Time-Training (TTT) is an approach to cope with out-of-distribution (OOD) data by adapting a trained model to distribution shifts occurring at test-time. We propose to perform this adaptation via Activation Matching (ActMAD): We analyze activations of the model and align activation statistics of the OOD test data to those of the training data. In contrast to existing methods, which model the distribution of entire channels in the ultimate layer of the feature extractor, we model the distribution of each feature in multiple layers across the network. This results in a more fine-grained supervision and makes ActMAD attain state of the art performance on CIFAR-100C and Imagenet-C. ActMAD is also architecture- and task-agnostic, which lets us go beyond image classification, and score 15.4% improvement over previous approaches when evaluating a KITTI-trained object detector on KITTI-Fog. Our experiments highlight that ActMAD can be applied to online adaptation in realistic scenarios, requiring little data to attain its full performance.