OCGAN: One-Class Novelty Detection Using GANs With Constrained Latent Representations

Abstract

We present a novel model called OCGAN for the classical problem of one-class novelty detection, where, given a set of examples from a particular class, the goal is to determine if a query example is from the same class. Our solution is based on learning latent representations of in-class examples using a denoising auto-encoder network. The key contribution of our work is our proposal to explicitly constrain the latent space to exclusively represent the given class. In order to accomplish this goal, firstly, we force the latent space to have bounded support by introducing a tanh activation in the encoder's output layer. Secondly, using a discriminator in the latent space that is trained adversarially, we ensure that encoded representations of in-class examples resemble uniform random samples drawn from the same bounded space. Thirdly, using a second adversarial discriminator in the input space, we ensure all randomly drawn latent samples generate examples that look real. Finally, we introduce a gradient-descent based sampling technique that explores points in the latent space that generate potential out-of-class examples, which are fed back to the network to further train it to generate in-class examples from those points. The effectiveness of the proposed method is measured across four publicly available datasets using two one-class novelty detection protocols where we achieve state-of-the-art results.

Perceptual Image Anomaly Detection

“…GANs [6] created a new branch in the development of image anomaly detection. GAN-based approaches [7][8][9][10][11] differ in two parts: (i) how to find latent vectors that correspond to the input images, (ii) how to estimate abnormality based Figure 2: Comparison of four anomaly detection models. G denotes the generator, E the encoder, D * the discriminators, "rec.…”

“…As shallow baselines we consider standard methods such as OC-SVM [16] and KDE [15]. We also test the performance of our approach against four state-of-the-art GAN-based methods: AnoGAN [7], ADGAN [8], OCGAN [9] and ALAD [11]. Finally, we report the performance of three deep learning approaches from different paradigms: Deep SVDD [22], GPND [20], and the Latent Space Autoregression approach [21] (results will be reported under the name LSA).…”

“…For ADGAN [8], OCGAN [9], ALAD [11], GPND [20], LSA [21] we used results as reported in the corresponding publications. Results for OC-SVM, KDE, AnoGAN were obtained from [21].…”

“…Evaluation Protocol. To test the methods on classification datasets, we use a one-vs-all evaluation scheme, which has recently been increasingly used in anomaly detection papers [8,9,11,[20][21][22]: to simulate out-of-distribution condition, one class of a dataset is considered as normal data while images of other classes are considered as abnormal. We evaluate results quantitatively using the area under the ROC curve (ROC AUC), which is a standard metric for this task.…”

“…Thus after training, the generator performs a mapping of the latent distribution to the data distribution. This property has been used [7][8][9] to estimate the likelihood of abnormality for an input: if there is a vector in latent space, which after passing through the generator could reconstruct the input object, the object is normal, otherwise it is not. The difference between an input and its closest reconstruction (reconstruction error) is used as an anomaly score for this object.…”