Hybrid Graph Neural Networks for Crowd Counting

Luo, Ao; Yang, Fan; Li, Xin; Nie, Dong; Jiao, Zhicheng; Zhou, Shangchen; Cheng, Hong

doi:10.1609/aaai.v34i07.6839

Cited by 75 publications

(29 citation statements)

References 27 publications

(44 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Crowd management systems have been developed to support services and infrastructures devoted to managing and controlling crowds at any time, so that in case of emergency situations, crowds are well managed, while the dangers and risks are minimized. Current studies focus on crowd counting and monitoring models [ 10 ] and algorithms [ 11 ], and crowd flow prediction architectures [ 12 ], their aim is to provide the government and local authorities with valuable information on large crowds [ 13 ].…”

Section: Smart Territory Platforms and The Edge Computing Approachmentioning

confidence: 99%

Deepint.net: A Rapid Deployment Platform for Smart Territories

Corchado

Chamoso

Hernández³

et al. 2021

Sensors

View full text Add to dashboard Cite

This paper presents an efficient cyberphysical platform for the smart management of smart territories. It is efficient because it facilitates the implementation of data acquisition and data management methods, as well as data representation and dashboard configuration. The platform allows for the use of any type of data source, ranging from the measurements of a multi-functional IoT sensing devices to relational and non-relational databases. It is also smart because it incorporates a complete artificial intelligence suit for data analysis; it includes techniques for data classification, clustering, forecasting, optimization, visualization, etc. It is also compatible with the edge computing concept, allowing for the distribution of intelligence and the use of intelligent sensors. The concept of smart cities is evolving and adapting to new applications; the trend to create intelligent neighbourhoods, districts or territories is becoming increasingly popular, as opposed to the previous approach of managing an entire megacity. In this paper, the platform is presented, and its architecture and functionalities are described. Moreover, its operation has been validated in a case study where the bike renting service of Paris—Vélib’ Métropole has been managed. This platform could enable smart territories to develop adapted knowledge management systems, adapt them to new requirements and to use multiple types of data, and execute efficient computational and artificial intelligence algorithms. The platform optimizes the decisions taken by human experts through explainable artificial intelligence models that obtain data from IoT sensors, databases, the Internet, etc. The global intelligence of the platform could potentially coordinate its decision-making processes with intelligent nodes installed in the edge, which would use the most advanced data processing techniques.

show abstract

Section: Smart Territory Platforms and The Edge Computing Approachmentioning

confidence: 99%

Deepint.net: A Rapid Deployment Platform for Smart Territories

Corchado

Chamoso

Hernández³

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…Early crowd counting networks typically employed multi-column structures [ 10 , 11 , 14 , 16 , 34 ] to model different scales of crowds. More recently, a graph network [ 35 ] was introduced to enhance scale-aware features. Perspective information of crowd scenes was also employed for networks [ 36 , 37 ] for improving the final counting performance.…”

Section: Related Workmentioning

confidence: 99%

Congested Crowd Counting via Adaptive Multi-Scale Context Learning

Zhang

Zhao

Duan

et al. 2021

Sensors

View full text Add to dashboard Cite

In this paper, we propose a novel congested crowd counting network for crowd density estimation, i.e., the Adaptive Multi-scale Context Aggregation Network (MSCANet). MSCANet efficiently leverages the spatial context information to accomplish crowd density estimation in a complicated crowd scene. To achieve this, a multi-scale context learning block, called the Multi-scale Context Aggregation module (MSCA), is proposed to first extract different scale information and then adaptively aggregate it to capture the full scale of the crowd. Employing multiple MSCAs in a cascaded manner, the MSCANet can deeply utilize the spatial context information and modulate preliminary features into more distinguishing and scale-sensitive features, which are finally applied to a 1 × 1 convolution operation to obtain the crowd density results. Extensive experiments on three challenging crowd counting benchmarks showed that our model yielded compelling performance against the other state-of-the-art methods. To thoroughly prove the generality of MSCANet, we extend our method to two relevant tasks: crowd localization and remote sensing object counting. The extension experiment results also confirmed the effectiveness of MSCANet.

show abstract

“…Recently, a lot of CNN-based methods have been proposed and advanced the performance of crowd counting. Most of them mainly solve various challenges of crowd counting in a fullysupervised manner, including large scale variations [4], [5], [24], [25], [26], [27], [28], [29], [30], attentive feature extraction [31], [32], [33], [34], [35], label noises [6], [7], empirical gaussian kernel [36], [37], [38], estimation uncertainty [39], [40], structural constraints [8], [9], [41], [42], and etc. These methods require a great number of labeled data in the training process which are rather burdensome for crowd counting.…”

Section: Crowd Countingmentioning

confidence: 99%

Reducing Spatial Labeling Redundancy for Semi-supervised Crowd Counting

Liu

Ren²,

Chai³

et al. 2021

Preprint

View full text Add to dashboard Cite

Labeling is onerous for crowd counting as it should annotate each individual in crowd images. Recently, several methods have been proposed for semi-supervised crowd counting to reduce the labeling efforts. Given a limited labeling budget, they typically select a few crowd images and densely label all individuals in each of them. Despite the promising results, we argue the None-or-All labeling strategy is suboptimal as the densely labeled individuals in each crowd image usually appear similar while the massive unlabeled crowd images may contain entirely diverse individuals. To this end, we propose to break the labeling chain of previous methods and make the first attempt to reduce spatial labeling redundancy for semi-supervised crowd counting. First, instead of annotating all the regions in each crowd image, we propose to annotate the representative ones only. We analyze the region representativeness from both vertical and horizontal directions, and formulate them as cluster centers of Gaussian Mixture Models. Additionally, to leverage the rich unlabeled regions, we exploit the similarities among individuals in each crowd image to directly supervise the unlabeled regions via feature propagation instead of the error-prone label propagation employed in the previous methods. In this way, we can transfer the original spatial labeling redundancy caused by individual similarities to effective supervision signals on the unlabeled regions. Extensive experiments on the widely-used benchmarks demonstrate that our method can outperform previous best approaches by a large margin.

show abstract

Hybrid Graph Neural Networks for Crowd Counting

Cited by 75 publications

References 27 publications

Deepint.net: A Rapid Deployment Platform for Smart Territories

Deepint.net: A Rapid Deployment Platform for Smart Territories

Congested Crowd Counting via Adaptive Multi-Scale Context Learning

Reducing Spatial Labeling Redundancy for Semi-supervised Crowd Counting

Contact Info

Product

Resources

About