Learning Memory-Guided Normality for Anomaly Detection

By Hyunjong Park, Jongyoun Noh, Bumsub Ham. Computer Vision and Pattern Recognition, 2020

894 citations recorded by the source corpus.

We address the problem of anomaly detection, that is, detecting anomalous events in a video sequence. Anomaly detection methods based on convolutional neural networks (CNNs) typically leverage proxy tasks, such as reconstructing input video frames, to learn models describing normality without seeing anomalous samples at training time, and quantify the extent of abnormalities using the reconstruction error at test time. The main drawbacks of these approaches are that they do not consider the diversity of normal patterns explicitly, and the powerful representation capacity of CNNs allows to reconstruct abnormal video frames. To address this problem, we present an unsupervised learning approach to anomaly detection that considers the diversity of normal patterns explicitly, while lessening the representation capacity of CNNs. To this end, we propose to use a memory module with a new update scheme where items in the memory record prototypical patterns of normal data. We also present novel feature compactness and separateness losses to train the memory, boosting the discriminative power of both memory items and deeply learned features from normal data. Experimental results on standard benchmarks demonstrate the effectiveness and efficiency of our approach, which outperforms the state of the art.

Subfields: Artificial Intelligence, Computer Networks and Communications, Epidemiology

View Paper: https://www.semanticscholar.org/paper/18f207d8dab7357f4f674211ec4f150de1c93a0e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Networks and Communications

In Epidemiology

Citations in the corpus

Anomaly Detection in Video via Self-Supervised and Multi-Task Learning (2020, methodology): “In recent years, a growing interest has been dedicated to the task of detecting anomalous events in video [8, 9, 10, 13, 17, 19, 20, 24, 30, 34, 35, 36, 37, 38, 39, 49, 51, 55, 57, 61, 62, 63].”
A Hybrid Video Anomaly Detection Framework via Memory-Augmented Flow Reconstruction and Flow-Guided Frame Prediction (2021, methodology): “From top to bottom, we show the sampled video frames, ground-truth abnormal sections (green regions are abnormal), result of MNAD-R [38], result of MNAD-P [38], result of VEC-VAD [47] and result of HF(2)-VAD.”
EfficientAD: Accurate Visual Anomaly Detection at Millisecond-Level Latencies (2023, methodology): “Recent autoencoder-based methods rely on accurate reconstructions of normal images and inaccurate reconstructions of anomalous images [8, 12, 20, 43].”
Spatio-Temporal Meta-Graph Learning for Traffic Forecasting (2022, neutral): “…distinguishing power for diverse scenarios on different roads over time, we regulate the memory parameters with two constraints (Gong et al. 2019; Park, Noh, and Ham 2020), including a contrastive loss L 1 and a consistency loss L 1 , denoted by: (9) where T denotes the total number of sequences…”
Anomaly Detection via Reverse Distillation from One-Class Embedding (2022, methodology): “In the context of one-class distillation for unsupervised AD, the student model is expected to generate highly different representations from the teacher when the queries are anomalous samples [11, 26].”
A Unified Model for Multi-class Anomaly Detection (2022, background): “2) The cross-attention between two sets of inputs is changed to a concatenation along the channel dimension and a linear projection, followed by layer normalization and ReLU activation.”
Generalized Out-of-Distribution Detection: A Survey (2021, neutral): “AE/VAE-based Models Apart from the standard combination of reconstruction-error and AE/VAE models [236], [237], other methods use more sophisticated strategies such as reconstructing by memorized normality [240], [241], adapting model architectures [242], and partial/conditional reconstruction [159]”
Weakly-supervised Video Anomaly Detection with Robust Temporal Feature Magnitude Learning (2021, methodology): “Our method RTFM achieves superior performance when compared with previous SOTA unsupervised learning methods [15, 27, 30, 41, 70] and weaklysupervised approaches [62, 74, 78].”