Plug-and-Play CNN for Crowd Motion Analysis: An Application in Abnormal Event Detection

Abstract

Most of the crowd abnormal event detection methods rely on complex hand-crafted features to represent the crowd motion and appearance. Convolutional Neural Networks (CNN) have shown to be a powerful instrument with excellent representational capacities, which can leverage the need for hand-crafted features. In this paper, we show that keeping track of the changes in the CNN feature across time can be used to effectively detect local anomalies. Specifically, we propose to measure local abnormality by combining semantic information (inherited from existing CNN models) with low-level optical-flow. One of the advantages of this method is that it can be used without the fine-tuning phase. The proposed method is validated on challenging abnormality detection datasets and the results show the superiority of our approach compared with the state-of-theart methods.

Robust and Computationally-Efficient Anomaly Detection Using Powers-Of-Two Networks

“…Most anomaly detection methods are based on motion information which use hand-crafted features to model normal-activity patterns [7][8][9][10][11][12][13][14][15][16]. On the other hand our method uses the entire set of motion vectors obtained from the video as in [17], [18], [19]. We also found the optical flow domain as a reasonable representation for training and evaluating our model.…”

SiTGRU: Single-Tunnelled Gated Recurrent Unit for Abnormality Detection

“…However, it is still a challenging problem due to two key factors. (i) The lack of large-scale annotated training data limits the performance of Deep Convolutional Neural Networks (DCNNs) [28,29]. (ii) The contentious definition of the term "abnormal" or "anomaly" causes significantly different solutions in different context (environment) [12,28,29].…”

“…(ii) The contentious definition of the term "abnormal" or "anomaly" causes significantly different solutions in different context (environment) [12,28,29]. Abnormality is not clearly defined as it is context-specific, leaving a room for unintentional subjectivity in data annotation [12,17,28]. For instance, Sun et al [34] de-fined "anomaly" as an event or scene that is rarely manifested.…”

“…Due to the robust and flexible underlying representation power [17,28], DCNNs are capable of classifying or regressing concepts by modelling them with large-scale positive samples and can better represent spatial and temporal features. Spatio-temporal based networks have shown their success in learning and representing data with spatial and temporal dimension like videos.…”

Everything You Wanted to Know about Deep Learning for Computer Vision but Were Afraid to Ask

“…Convolutional Neural Networks (CNNs), Deep Belief Nets (DBNs), Restricted Boltzmann Machines (RBMs) and Autoencoders (AEs), started appearing as a basis for state-of-the-art methods in several computer vision applications (e.g. remote sensing [7], surveillance [8], [9] and re-identification [10]). The ImageNet challenge [1] played a major role in this process, starting a race for the model that could beat the current champion in the image classification challenge, but also image segmentation, object recognition and other tasks.…”