Energy-Based Anomaly Detection and Localization
arXiv:2105.03270
Abstract
This brief sketches initial progress towards a unified energy-based solution for the semi-supervised visual anomaly detection and localization problem. In this setup, we have access to only anomaly-free training data and want to detect and identify anomalies of an arbitrary nature on test data. We employ the density estimates from the energy-based model (EBM) as normalcy scores that can be used to discriminate normal images from anomalous ones. Further, we back-propagate the gradients of the energy score with respect to the image in order to generate a gradient map that provides pixel-level spatial localization of the anomalies in the image. In addition to the spatial localization, we show that simple processing of the gradient map can also provide alternative normalcy scores that either match or surpass the detection performance obtained with the energy value. To quantitatively validate the performance of the proposed method, we conduct experiments on the MVTec industrial dataset. Though still preliminary, our results are very promising and reveal the potential of EBMs for simultaneously detecting and localizing unforeseen anomalies in images.
arXiv comment9 pages, 3 figures, as submitted to EBM ICLR 2021 workshop
References in corpus (10)
- Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
- A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
- Improving Unsupervised Defect Segmentation by Applying Structural Similarity to Autoencoders
- Deep Anomaly Detection with Outlier Exposure
- Sub-Image Anomaly Detection with Deep Pyramid Correspondences
- Deep Structured Energy Based Models for Anomaly Detection
- Implicit Generation and Generalization in Energy-Based Models
- Your Classifier is Secretly an Energy Based Model and You Should Treat it Like One
- Patch SVDD: Patch-level SVDD for Anomaly Detection and Segmentation
- Probabilistic Modeling of Deep Features for Out-of-Distribution and Adversarial Detection