Unmasking the Abnormal Events in Video

By Radu Tudor Ionescu, Sorina Smeureanu, Bogdan Alexe, Marius Claudiu Popescu. IEEE International Conference on Computer Vision, 2017

310 citations recorded by the source corpus.

We propose a novel framework for abnormal event detection in video that requires no training sequences. Our framework is based on unmasking, a technique previously used for authorship verification in text documents, which we adapt to our task. We iteratively train a binary classifier to distinguish between two consecutive video sequences while removing at each step the most discriminant features. Higher training accuracy rates of the intermediately obtained classifiers represent abnormal events. To the best of our knowledge, this is the first work to apply unmasking for a computer vision task. We compare our method with several state-of-the-art supervised and unsupervised methods on four benchmark data sets. The empirical results indicate that our abnormal event detection framework can achieve state-of-the-art results, while running in real-time at 20 frames per second.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Epidemiology

View Paper: https://www.semanticscholar.org/paper/a4a0a1e2b573affc16de38b7fff91f6e2507140b

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 Vision and Pattern Recognition

In Epidemiology

Citations in the corpus

Deep Learning for Anomaly Detection (2020, methodology): “One interesting work of this line is the unmasking framework for online anomaly detection [66].”
Latent Space Autoregression for Novelty Detection (2018, unknown): “Concerning ShanghaiTech, we removed the dependency on UCSD Ped2 ShanghaiTech MPPCA [16] 0.693 - MPPC+SFA [28] 0.613 - MDT [28] 0.829 - ConvAE [11] 0.850 0.609 ConvLSTM-AE [26] 0.881 - Unmasking [14] 0.822 - Hinami et al .”
Self-Supervised Predictive Convolutional Attentive Block for Anomaly Detection (2021, methodology): “Anomaly detection methods can be classified into: dictionary learning methods [7–9, 14, 40, 55], change detection frameworks [11, 26, 38, 48], probability-based methods [1, 2, 16, 23, 29, 44, 45, 56, 61, 74], distance-based models [6, 10, 25, 27, 50, 51,53,57,58,63,65,68,70], and reconstruction-base”
Future Frame Prediction for Anomaly Detection - A New Baseline (2017, methodology): “We also report the running time of other methods such as 20 fps in [31], 150 fps [20] and 0.”
Object-Centric Auto-Encoders and Dummy Anomalies for Abnormal Event Detection in Video (2018, methodology): “Compared to most of the recent works [13, 21, 22, 24, 33], our method provides an absolute gain of more than 5% in terms of frame-level AUC.”
Anomaly Detection in Video Sequence With Appearance-Motion Correspondence (2019, unknown): “Some researchers apply pretrained classification models (such as VGG [41]) to extract useful features from input videos [42, 16].”
Learning Memory-Guided Normality for Anomaly Detection (2020, methodology): “, 20 fps for Unmasking [41], 50 fps for StackRNN [26], 25 fps for Frame-Pred [22], and 45 fps for MemAE [8] with the same setting as ours.”
Anomaly Detection in Video via Self-Supervised and Multi-Task Learning (2020, methodology): “Frame-level AUC scores (in %) of the state-of-the-art methods [7, 8, 9, 10, 13, 14, 16, 17, 18, 19, 20, 21, 23, 24, 27, 28, 29, 30, 31, 32, 33, 34, 36, 37, 38, 40, 41, 47, 48, 49, 51, 53, 55, 57, 59, 60, 61, 62, 64] versus our deep+wide architecture trained on four proxy tasks at the object level, a”
Weakly-supervised Video Anomaly Detection with Robust Temporal Feature Magnitude Learning (2021, methodology): “With the advent of deep learning, more recent approaches use the features from pre-trained deep neural networks [11, 19, 40, 54, 77].”
Memorizing Normality to Detect Anomaly: Memory-Augmented Deep Autoencoder for Unsupervised Anomaly Detection (2019, unknown): “UCSD-Ped2 CUHK ShanghaiTech MPPCA [14] 0.693 - - MPPCA+SFA [27] 0.613 - - MDT [27] 0.829 - - Unmasking [37] 0.822 0.806 - MT-FRCN [11] 0.922 - - AE-Conv2D [10] 0.850 0.800 0.609 AE-Conv3D [45] 0.912 0.771 - TSC [26] 0.910 0.806 0.679 StackRNN [26] 0 where e u denotes the reconstruction error the u…”
Part-Activated Deep Reinforcement Learning for Action Prediction (2018, background): “As an important branch of human activity analysis, predicting the activity of humans presents importance in a number of real-world applications, such as video detection [43], abnormal behavior detection [7, 40] and robot interaction [36].”