Qatar College develops system to handle crowd throughout FIFA World Cup

Doha: In collaboration with the Supreme Committee for Supply and Legacy (SC), Qatar College (QU) School of Engineering has developed an clever crowd administration and management programs, together with a number of parts for crowd counting, face recognition, and irregular occasion detection (AED). 

The QU analysis group, led by Prof. Sumaya Al Maadeed because the lead principal investigator within the examine, consists of Dr. Noor Al Maadeed, Affiliate Dean of Graduate Research for Tutorial Affairs and Affiliate Professor of Pc Engineering, Dr. Khalid Abualsaud, Lecturer of Pc Engineering, Prof. Amr Mohamed, Professor of Pc Engineering, Prof. Tamer Khattab, Professor {of electrical} engineering and Performing Director of Excellence in Educating and Studying Middle, Dr. Yassine Himeur, Submit-doctoral Researcher, and Dr. Omar Elharrouss. Submit-doctoral Researcher, Najmath Ottakath, Grasp’s scholar.

The safety and security of gamers, spectators, and others related to the FIFA World Cup Qatar 2022, is on the centre of the eye of the organising committee. Usually, safety dangers improve multi-fold, contemplating the massive scale of the occasion and the numerous variety of followers anticipated to attend (greater than 1.5 million followers). Thus, the safety of the FIFA World Cup Qatar 2022 is difficult because of the rising variety of doable threats and use of expertise.

Crowd administration on the World Cup stadiums and their perimeters is essential to make sure the protection and smoothness of the World Cup occasions because of the inherent occlusion and density of the gang inside and out of doors the stadiums. Qatar 2022 will depend on the deployment of cutting-edge applied sciences, comparable to surveillance drones, ICT, and AI, to optimise crowd administration. 

On this respect, the QU analysis group has first developed a crowd counting system from drones’ information, which exploits the dilated and scaled neural networks to extract pertinent options and density crowd estimations. 

Moreover, a brand new dataset for crowd counting in sports activities services named Soccer Supporters Crowd Dataset (FSC-Set) is launched. It consists of 6000 photographs labeled manually and representing varied forms of scenes, containing hundreds of individuals gathering in or across the stadiums. 

The analysis group’s effort has additionally targeted on growing a face recognition system, which considers faces underneath pose variations utilizing a multitask convolutional neural community (CNN). Particularly, a cascade construction was employed to mix a pose estimation strategy and a face identification module. The CNN-based pose estimation strategy has been skilled on three classes of face photographs, together with left aspect, frontal, and proper aspect captures. 

Shifting on, three CNN structure, particularly VGG-16+PReLU left, VGG-16+PReLU entrance, and VGG-16+PReLU proper, have been deployed to establish faces primarily based on the estimated pose. Moreover, a skin-based face segmentation scheme, primarily based on structure-texture decomposition and a color-invariant description, has been launched to take away ineffective face data (e.g., background content material). Empirical evaluations have been carried out on 4 widespread face recognition datasets, the place the proposed system has outperformed associated state-of-the-art schemes. 

Lately, utilizing drone-based video surveillance, irregular occasion detection (AED) is receiving rising consideration attributable to its reliability and cost-effectiveness. Usually, drones augmented with cameras can spot violent behaviors in crowds throughout sports activities occasions. They’ll monitor crowds within the perimeter of stadiums and/or different public venues through the World Cup. 

To that finish, the analysis group, led by Prof. Al Maadeed, has developed a novel AED, which goals at studying irregular actions utilizing each regular and irregular segments. It allows to keep away from the annotation of anomalous occasions in coaching video sequences to cut back the computational value and therefore be simply applied on drones. Due to this fact, irregular occasions are discovered utilizing a deep a number of occasion rating scheme, which leverages weakly annotated coaching video sequences. Put merely, coaching annotations are placed on entire movies as an alternative of particular clips.



Leave a Reply

Your email address will not be published.