Where Are They Going? Predicting Human Behaviors in Crowded Scenes

Abstract

In this article, we propose a framework for crowd behavior prediction in complicated scenarios. The fundamental framework is designed using the standard encoder-decoder scheme, which is built upon the long short-term memory module to capture the temporal evolution of crowd behaviors. To model interactions among humans and environments, we embed both the social and the physical attention mechanisms into the long short-term memory. The social attention component can model the interactions among different pedestrians, whereas the physical attention component helps to understand the spatial configurations of the scene. Since pedestrians’ behaviors demonstrate multi-modal properties, we use the generative model to produce multiple acceptable future paths. The proposed framework not only predicts an individual’s trajectory accurately but also forecasts the ongoing group behaviors by leveraging on the coherent filtering approach. Experiments are carried out on the standard crowd benchmarks (namely, the ETH, the UCY, the CUHK crowd, and the CrowdFlow datasets), which demonstrate that the proposed framework is effective in forecasting crowd behaviors in complex scenarios.

Early warning on safety risk of highly aggregated tourist crowds based on VGGT-Count network model

“…Alabdulkarim et al [23] defined crowd management as the practice of controlling crowd activities before, during, and after events, including handling all elements such as personnel, venues, facilities, data, and technology. In terms of management strategies, scholars have studied various aspects such as sociology, psychology, and computer science, including crowd evacuation [24], crowd behavior [25][26][27][28], and crowd modeling [29][30][31][32]. Traditional crowd management strategies need to be integrated with technological means to provide accurate crowd-related information for optimal management [33].…”

Bidirectional Transformer GAN for Long-term Human Motion Prediction

“…Although these RNN-based approaches performed an interesting exploration, one can still observe unsatisfactory aspects in the predicted motion sequences. In order to ix these limitations, several works use feed-forward networks other than RNNs to model human pose [3,23,26,33,34,40,48]. For example, Butepage et al [3] proposed a deep learning fully-connected network that investigates diferent strategies to encode temporal, and historical information and generalizes well to new, unseen motions.…”

PMAL: A Proxy Model Active Learning Approach for Vision Based Industrial Applications

“…Since these patches retains no information about the position, therefore, positional information is added in the form of as demonstrated in Figure 3(a). The final sequence of patches with token achieved because of these operations is given in the Equation (4).…”

“…Data acts as fuel for Deep Learning (DL) models and in today's era, the availability of huge amount of multimedia data enable DL models to be applied in various realms of life [1,2]. Different industrial domains such as surveillance [3,4], medical sciences [5,6], remote sensing [7], disaster management [8], defense [9], transportation [10], and entertainment [11,12] have been tremendously flourished with the advancements in DL [13]. However, the acquisition of properly annotated data for training is deemed as a substantial challenge to the wider adoption of DL models in the industry.…”