Deep Generative Model using Unregularized Score for Anomaly Detection with Heterogeneous Complexity
This work addresses anomaly detection in manufacturing and natural image datasets, but it is incremental as it modifies regularization terms in existing deep generative models.
The paper tackled the problem of anomaly detection in images with heterogeneous complexity, where existing probabilistic models fail on simple objects with small anomalies. The authors proposed an unregularized score for deep generative models, which demonstrated robustness to sample complexity and improved anomaly detection on toy and real-world manufacturing datasets.
Accurate and automated detection of anomalous samples in a natural image dataset can be accomplished with a probabilistic model for end-to-end modeling of images. Such images have heterogeneous complexity, however, and a probabilistic model overlooks simply shaped objects with small anomalies. This is because the probabilistic model assigns undesirably lower likelihoods to complexly shaped objects that are nevertheless consistent with set standards. To overcome this difficulty, we propose an unregularized score for deep generative models (DGMs), which are generative models leveraging deep neural networks. We found that the regularization terms of the DGMs considerably influence the anomaly score depending on the complexity of the samples. By removing these terms, we obtain an unregularized score, which we evaluated on a toy dataset and real-world manufacturing datasets. Empirical results demonstrate that the unregularized score is robust to the inherent complexity of samples and can be used to better detect anomalies.