Anomaly Detection in Video Sequence With Appearance-Motion Correspondence

Nguyen, Trong-Nguyen; Meunier, Jean

doi:10.1109/iccv.2019.00136

Cited by 355 publications

(172 citation statements)

References 46 publications

Supporting

Mentioning

172

Contrasting

Order By: Relevance

“…Nguyen et al [2] propose a model that is designed as a combination of a network for reconstruction and a model for image translation, which shares the same encoder. The former sub-network identifies the most significant structures that appear in video frames, and the latter attempts to connect movement models to those structures.…”

Section: Related Workmentioning

confidence: 99%

Violence Detection In Surveillance Videos Using Deep Learning

Moaaz

Mohamed

2020

النشرة المعلوماتیة فی الحاسبات والمعلومات

View full text Add to dashboard Cite

Nowadays computer technologies are flowering especially the artificial intelligence field. It lives its prosperous years. Recently it closes the gap between humans and machines with the facilitation of supporting decisions. One of these gaps is the surveillance cameras labors' attentiveness and the lack of instantaneous detection of violence actions on the scenes of such cameras. In this paper we present an end to end deep neural network to detect the violence scenes in the surveillance cameras, the proposed system composed of set of phases. It extracts a set of selectively distributed frames of the video clip, performs spatio-temporal features, and passes them to a fully connected neural to classify the video to violence or non-violence action. The model is evaluated on different datasets; like Real Life Violence Situations aka RLVS and Hockey Fight Detection datasets. The accuracy was 92% and 94.5% respectively, which outperformed the previous related works.

show abstract

Section: Related Workmentioning

confidence: 99%

Violence Detection In Surveillance Videos Using Deep Learning

Moaaz

Mohamed

2020

النشرة المعلوماتیة فی الحاسبات والمعلومات

View full text Add to dashboard Cite

show abstract

“…Most of recently successful studies [ 1 , 2 , 3 , 4 , 5 , 6 ] have tackled this challenge in specific unsupervised ways. They only use normal samples from training sets to generate the standards for normal actions, then try to enlarge the deviation between abnormal samples in the test set and their standards.…”

Section: Introductionmentioning

confidence: 99%

“…The state of the art proposes two approaches for defining the anomalous feature that are, respectively, based on changing detection [ 1 , 3 , 7 ] and reconstruction/prediction errors [ 4 , 5 , 6 ]. The first solution is a natural approach where each event is compared with its neighbors to find the most different ones.…”

Section: Introductionmentioning

confidence: 99%

“…In terms of feature extractions and descriptors encoding, previous anomaly detection frameworks proposed various feature types reasoning: at object-centric or local level [ 2 , 6 , 9 , 10 , 11 ], at frame level [ 4 , 5 ], or at both [ 1 ]. On the one hand, some methods encode their features to calculate abnormality scores; the decision is done thanks to a threshold or a peak estimation [ 4 , 5 ]. In fact, by integrating all features into one score value, spatial and temporal information are discarded, and a simple comparing decision model is learned.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Multi-Channel Generative Framework and Supervised Learning for Anomaly Detection in Surveillance Videos

Boonært

Ambellouis

et al. 2021

Sensors

View full text Add to dashboard Cite

Recently, most state-of-the-art anomaly detection methods are based on apparent motion and appearance reconstruction networks and use error estimation between generated and real information as detection features. These approaches achieve promising results by only using normal samples for training steps. In this paper, our contributions are two-fold. On the one hand, we propose a flexible multi-channel framework to generate multi-type frame-level features. On the other hand, we study how it is possible to improve the detection performance by supervised learning. The multi-channel framework is based on four Conditional GANs (CGANs) taking various type of appearance and motion information as input and producing prediction information as output. These CGANs provide a better feature space to represent the distinction between normal and abnormal events. Then, the difference between those generative and ground-truth information is encoded by Peak Signal-to-Noise Ratio (PSNR). We propose to classify those features in a classical supervised scenario by building a small training set with some abnormal samples of the original test set of the dataset. The binary Support Vector Machine (SVM) is applied for frame-level anomaly detection. Finally, we use Mask R-CNN as detector to perform object-centric anomaly localization. Our solution is largely evaluated on Avenue, Ped1, Ped2, and ShanghaiTech datasets. Our experiment results demonstrate that PSNR features combined with supervised SVM are better than error maps computed by previous methods. We achieve state-of-the-art performance for frame-level AUC on Ped1 and ShanghaiTech. Especially, for the most challenging Shanghaitech dataset, a supervised training model outperforms up to 9% the state-of-the-art an unsupervised strategy.

show abstract

“…Anomaly detection is a hot topic in various domains, such as manufacturing anomaly detection [ 1 , 2 ], cyber attack detection [ 3 , 4 , 5 , 6 ], and crowded scene video anomaly detection [ 7 , 8 ]. Cyber attacks detection typically detects three types of external attacks, i.e., false data injection attack, denial-of-service attack, and confidentiality attack.…”

Section: Introductionmentioning

confidence: 99%

Anomaly Detection in Discrete Manufacturing Systems by Pattern Relation Table Approaches

Sun

Zhen

2020

Sensors

View full text Add to dashboard Cite

Anomaly detection for discrete manufacturing systems is important in intelligent manufacturing. In this paper, we address the problem of anomaly detection for the discrete manufacturing systems with complicated processes, including parallel processes, loop processes, and/or parallel with nested loop sub-processes. Such systems can generate a series of discrete event data during normal operations. Existing methods that deal with the discrete sequence data may not be efficient for the discrete manufacturing systems or methods that are dealing with manufacturing systems only focus on some specific systems. In this paper, we take the middle way and seek to propose an efficient algorithm by applying only the system structure information. Motivated by the system structure information that the loop processes may result in repeated events, we propose two algorithms—centralized pattern relation table algorithm and parallel pattern relation table algorithm—to build one or multiple relation tables between loop pattern elements and individual events. The effectiveness of the proposed algorithms is tested by two artificial data sets that are generated by Timed Petri Nets. The experimental results show that the proposed algorithms can achieve higher AUC and F1-score, even with smaller sized data set compared to the other algorithms and that the parallel algorithm achieves the highest performance with the smallest data set.

show abstract

Anomaly Detection in Video Sequence With Appearance-Motion Correspondence

Cited by 355 publications

References 46 publications

Violence Detection In Surveillance Videos Using Deep Learning

Violence Detection In Surveillance Videos Using Deep Learning

Multi-Channel Generative Framework and Supervised Learning for Anomaly Detection in Surveillance Videos

Anomaly Detection in Discrete Manufacturing Systems by Pattern Relation Table Approaches

Contact Info

Product

Resources

About