A Deep One-Class Neural Network for Anomalous Event Detection in Complex Scenes

A Unifying Review of Deep and Shallow Anomaly Detection

“…Deep one-class classification methods aim to overcome these challenges by learning useful neural network feature maps φω : X → Z from the data or transferring such networks from related tasks. Deep SVDD [137], [144], [145], [327] and deep OC-SVM variants [136], [328] (16) and 20, respectively. These methods are typically optimized with SGD variants [329]- [331], which, together with GPU parallelization, makes them scale to large data sets.…”

“…A recurring question in deep one-class classification is how to meaningfully regularize against a feature map collapse φω ≡ c. Without regularization, minimum volume or maximum margin objectives, such as (16), (20), or (22), could be trivially solved with a constant mapping [137], [333]. Possible solutions for this include adding a reconstruction term or architectural constraints [137], [327], freezing the embedding [136], [139], [140], [142], [334], inversely penalizing the embedding variance [335], using true [144], [336], auxiliary [139], [233], [332], [337], or artificial [337] negative examples in training, pseudolabeling [152], [153], [155], [335], or integrating some manifold assumption [333]. Further variants of deep one-class classification include multimodal [145] or time-series extensions [338] and methods that employ adversarial learning [138], [141], [339] or transfer learning [139], [142].…”

Positive unlabeled learning‐based anomaly detection in videos

“…Figure 7 shows examples of the pixel‐level detection results. The proposed method is compared with SCL, 30 semi‐supervised learning, 11 one‐class classification 17 and other PU learning 20 methods. It is inferred from Figure 7 that our method reaches successfully detects anomalies and can locate the anomalies more accurately than other methods.…”

“…(C) Detection results of TSR‐AE 11 . (D) Detection results of DOCNN 17 . (E) Detection results of RPNA 20 .…”

“…(iii) The reconstruction methods , 16 use the prior information of the data and formulate a hypothesis for generating process, a model is selected and build a classifier to detect abnormal data. In classification or regression problems, a more powerful Bayesian‐based method can be used to detect outliers 17,18 . Rather than using the most appropriate weights for a classifier or a regressor to calculate the output, the output is weighted adopting a probability that ensures the weights are correct.…”