Skip-GANomaly: Skip Connected and Adversarially Trained Encoder-Decoder Anomaly Detection

Abstract

Despite inherent ill-definition, anomaly detection is a research endeavour of great interest within machine learning and visual scene understanding alike. Most commonly, anomaly detection is considered as the detection of outliers within a given data distribution based on some measure of normality. The most significant challenge in real-world anomaly detection problems is that available data is highly imbalanced towards normality (i.e. non-anomalous) and contains a most a sub-set of all possible anomalous samples -hence limiting the use of well-established supervised learning methods. By contrast, we introduce an unsupervised anomaly detection model, trained only on the normal (non-anomalous, plentiful) samples in order to learn the normality distribution of the domain, and hence detect abnormality based on deviation from this model. Our proposed approach employs an encoder-decoder convolutional neural network with skip connections to thoroughly capture the multi-scale distribution of the normal data distribution in highdimensional image space. Furthermore, utilizing an adversarial training scheme for this chosen architecture provides superior reconstruction both within high-dimensional image space and a lower-dimensional latent vector space encoding. Minimizing the reconstruction error metric within both the image and hidden vector spaces during training aids the model to learn the distribution of normality as required. Higher reconstruction metrics during subsequent test and deployment are thus indicative of a deviation from this normal distribution, hence indicative of an anomaly. Experimentation over established anomaly detection benchmarks and challenging real-world datasets, within the context of X-ray security screening, shows the unique promise of such a proposed approach.

A GAN-Based Anomaly Detection Approach for Imbalanced Industrial Time Series

“…They first proposed a training algorithm for no-negative samples and achieved state-of-the-art performance for anomaly detecting in some image benchmark datasets. Motivated by [1] and [2], aimed at univariate time series data in industrial area, we propose a new network trained only on normal samples aimed at detecting fault in time series dataset from industrial area. We adopt the similar encoderdecoder-encoder three-sub-networks in generator, but with a different network architecture.…”

“…To solve the imbalanced industrial data for fault diagnosis and preprocess the huge amount of time series before training the model, motivated by [1], [2], we propose a novel GAN[3]based approach combing the advantages of feature extractor and GAN. The main contributions of this paper are as follows: 1) For the univariate time series in the industrial field, a fault detection algorithm based on GAN is proposed for the first time.…”

Building Visual Anomaly Dataset from Satellite Data using ADS-B

“…A standard approach is to use one of the standard object classification datasets, where one or more categories are arbitrarily designated to be anomalous. Such approach has been used in performance analysis of the GANomaly algorithm and its more recent derivative [2,3]. They used CIFAR10 [8] and MNIST [9] datasets in the manner described.…”

“…This can be solved by limiting our dataset to aforementioned one-class approaches, where algorithm is not supplied with anomaly examples during the training, such as [3]. In this case we are free to decide the nature of anomaly in the dataset, as the challenge lies in the good modelling of normal data, and detecting deviations from that.…”

“…Anomaly detection is one of the classical problems in artificial intelligence and computer vision [3]. The core task of anomaly detection is to find unusual patterns in data.…”