GODS: Generalized One-Class Discriminative Subspaces for Anomaly Detection

Abstract

One-class learning is the classic problem of fitting a model to data for which annotations are available only for a single class. In this paper, we propose a novel objective for one-class learning. Our key idea is to use a pair of orthonormal frames -as subspaces -to "sandwich" the labeled data via optimizing for two objectives jointly: i) minimize the distance between the origins of the two subspaces, and ii) to maximize the margin between the hyperplanes and the data, either subspace demanding the data to be in its positive and negative orthant respectively. Our proposed objective however leads to a non-convex optimization problem, to which we resort to Riemannian optimization schemes and derive an efficient conjugate gradient scheme on the Stiefel manifold.To study the effectiveness of our scheme, we propose a new dataset Dash-Cam-Pose, consisting of clips with skeleton poses of humans seated in a car, the task being to classify the clips as normal or abnormal; the latter is when any human pose is out-of-position with regard to say an airbag deployment. Our experiments on the proposed Dash-Cam-Pose dataset, as well as several other standard anomaly/novelty detection benchmarks demonstrate the benefits of our scheme, achieving state-of-the-art oneclass accuracy. 5 The data follows the formula f (x) = √ x * (x + sign(randn) * rand), where randn and rand are standard MATLAB functions.

From unsupervised to semi-supervised anomaly detection methods for HRRP targets

“…[2018] Du et al [2019] Wan et al [2019] Wang and Cherian [2019. For each experiment, over the 48000 HRRPs of our dataset a random fraction of 10% defines the test set.…”

“…Since we use the standard Gaussian (radial basis function) kernel in our experiments with OC-SVM, there was no point in adding SVDD to this study eventhough its name suggests a closer relation with one of our highlighted AD methods. Additionally, as mentioned in Wang and Cherian [2019], let us remind that SVDD makes a strong isotropy hypothesis regarding the latent distribution of data.…”

Weakly-supervised Joint Anomaly Detection and Classification

“…Due to its massive application in video surveillance and crime scene investigation several research studies [8], [3], [12], [14], [21] have been reported in the last decade using hand-crafted features. Many earlier approaches [17], [20], [2], [4], [6], [11], [15], [24] address the problem of anomaly detection with one-class learning methods. These methods learn only from the normal patterns during training, defined for a specific scene.…”

Weakly-supervised Joint Anomaly Detection and Classification

“…Due to its massive application in video surveillance and crime scene investigation several research studies [8], [3], [12], [14], [22] have been reported in the last decade using hand-crafted features. Many earlier approaches [18], [21], [2], [4], [6], [11], [15], [25] address the problem of anomaly detection with one-class learning methods. These methods learn only from the normal patterns during training, defined for a specific scene.…”