Anomaly Event Detection in Security Surveillance Using Two-Stream Based Model

Abstract

Anomaly event detection has been extensively researched in computer vision in recent years. Most conventional anomaly event detection methods can only leverage the single-modal cues and not deal with the complementary information underlying other modalities in videos. To address this issue, in this work, we propose a novel two-stream convolutional networks model for anomaly detection in surveillance videos. Specifically, the proposed model consists of RGB and Flow two-stream networks, in which the final anomaly event detection score is the fusion of those of two networks. Furthermore, we consider two fusion situations, including the fusion of two streams with the same or different number of layers respectively. The design insight is to leverage the information underlying each stream and the complementary cues of RGB and Flow two-stream sufficiently. Two datasets (UCF-Crime and ShanghaiTech) are used to validate the effectiveness of proposed solution.

Exploring Convolutional Recurrent architectures for anomaly detection in videos: a comparative study

“…For each of the models, the frame-wise losses are calculated using MSE and temporally aggregated per video. The aggregated loss e(t) at time t are used to calculate the regularity s(t) which denotes the probability of a frame being normal 7 . The temporal regularity per video s(t) is calculated using the Eq.…”

“…Understanding videos has been one of the most challenging and open problems in computer vision [1][2][3] for applications such as action recognition, scene description, video captioning, video summarization and video anomaly detection. Video Anomaly Detection (VAD) is the process of identifying abnormal, rare and novel events concerning time and region of the video frames with several real-world applications in areas like security, surveillance [4][5][6][7][8], manufacturing [9], medicine [10] etc. Deep learning and Convolutional Neural Networks are predominantly used for visual tasks owing to their superior performance which can be attributed to their ability to uncover and learn hidden patterns and generalize well on huge datasets.…”

Anomaly detection in surveillance videos based on H265 and deep learning

“…Hao et al [23] presented a two-stream convolutional network model that was unique. The proposed model consists of two-stream flow and (red-green-blue) RGB networks, The completed anomaly activity recognition score is determined by the combined score.…”

“…Although the scale of AUC has increased to the existing work in this field when using HEVC, in contrast with the results got from using the original data, the results were not satisfactory. [14] 75.41 Anala et al 2019 [18] 85 Liu and Ma 2019 [19] 82 Zhang et al 2019 [20] 78.66 Zhu and Newsam 2019 [21] 79 Zhong et al [22] 82.12 Shreyas et al 2020 [16] 79.8 Hao et al 2020 [23] 81. 22 Zaheer et al 2020 [26] 83.03% Dubey et al 2021 [27] 81.91 Ullah et al 2021 [3] 78.43 Ullah et al 2021 [1] 85.53 Zaheer et al 2021 [32] 78.27 Majhi et al 2021 [33] 82.12 Wu et al 2018 [34] 87.65 Tian et al 2021 [35] 84.30% Cao et al 2022 [36] 83…”

A Review on Detection of Abnormal Events Using Machine Learning Approach

“…To detect abnormal events, an unsupervised k-means clustering algorithm that separates all the similar activities into a group is applied and it takes out irregular, dissimilar, abnormal events as outputs that we expected [7]. For getting an intuition regarding the structure of the data, clustering is the most frequent exploratory technique used for data analysis [8]. It can be defined in terms of task for identifying subgroups in the data, like the data points in the very same subgroup (cluster) are more similar whereas the data points that are present in different clusters were very different [9].…”