PANDA: Perceptually Aware Neural Detection of Anomalies

Barker, J.W.; Breckon, T.P.

Abstract

Semi-supervised methods of anomaly detection have seen substantial advancement in recent years. Of particular interest are applications of such methods to diverse, real-world anomaly detection problems where anomalous variations can vary from the visually obvious to the very subtle. In this work, we propose a novel fine-grained VAE-GAN architecture trained in a semi-supervised manner in order to detect both visually distinct and subtle anomalies. With the use of a residually connected dual-feature extractor, a fine-grained discriminator and a perceptual loss function, we are able to detect subtle, low inter-class (anomaly vs. normal) variant anomalies with greater detection capability and smaller margins of deviation in AUC value during inference compared to prior work whilst also remaining time-efficient during inference. We achieve stateof-the-art anomaly detection results when compared extensively with prior semi-supervised approaches across a multitude of anomaly detection benchmark tasks including trivial leave-oneout tasks (CIFAR-10 - AUPRCavg: 0.91; MNIST - AUPRCavg: 0.90) in addition to challenging real-world anomaly detection tasks (plant leaf disease - AUC: 0.776; threat item X-ray - AUC: 0.51), video frame-level anomaly detection (UCSDPed1 - AUC: 0.95) and high frequency texture with object anomalous defect detection (MVTEC - AUCavg: 0.83).

Citation

Barker, J., & Breckon, T. (2021, July). PANDA: Perceptually Aware Neural Detection of Anomalies. Presented at 2021 International Joint Conference on Neural Networks (IJCNN), Shenzhen, China

• Presentation Conference Type: Conference Paper (published)
• Conference Name: 2021 International Joint Conference on Neural Networks (IJCNN)
• Start Date: Jul 18, 2021
• End Date: Jul 22, 2021
• Acceptance Date: Apr 10, 2021
• Online Publication Date: Sep 23, 2021
• Publication Date: 2021
• Deposit Date: Apr 28, 2021
• Publicly Available Date: Apr 30, 2021
• Publisher: Institute of Electrical and Electronics Engineers
• DOI: https://doi.org/10.1109/ijcnn52387.2021.9534399
• Public URL: https://durham-repository.worktribe.com/output/1140878