Automatic detection of people with reduced mobility using YOLOv5 and data reduction strategy
Context: A portion of the users in the São Paulo Metro are people who have physical limitations and need the help of wheelchairs or other similar devices. In this way, the Metro stations have elevators that allow these users to move between the floors of the station. In order, for the elevator to be used, it is necessary for the user to call the operators of the stations, who, in turn, check if the user who is requesting access to the elevator fits the target audience. Problem: This type of request requires manual validation by station operators, causing interruptions in their work routines and delays in passenger travel. Solution: To implement and evaluate artificial intelligence methods for automatic detection of people in wheelchairs or other auxiliary devices. IS Theory: This project was idealized from the perspective of Customer Focus Theory. Method: The You Only Look Once (YOLOv5) neural network was implemented in the Mobility Aids database. Tests were performed considering the original and modified base, composed of a reduced number of images, aiming to assess whether the accuracy of the model remains even with reduced database data. Summary of Results: The results obtained show an average accuracy of more than 92% with the modified database. Contribution: The results corroborated our methodology and we will be able to test in Sao Paulo subway with real images. In a long term, It is expected that by automating such a task, operators will be less overloaded and passengers with reduced mobility will gain more autonomy.
Alexey Bochkovskiy, Chien-Yao Wang, and Hong-Yuan Mark Liao. 2020. Yolov4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934 (2020).
Brasil. 1988. Constituição da República Federativa do Brasil de 1988. Diário Oficial da União. http://www.planalto.gov.br/ccivil_03/leis/l7853.html
Brasil. 1989. Lei n. 7.853, de 24 de Outubro de 1989. Dispõe sobre o apoio às pessoas portadoras de deficiência. Diário Oficial da União. http://www.planalto.gov.br/ccivil_03/leis/l7853.html.
Organização Mundial de Saúde. 2021. Assistive technology. https://www.who.int/news-room/fact-sheets/detail/assistive-technology
Jun Deng, Xiaojing Xuan, Weifeng Wang, Zhao Li, Hanwen Yao, and Zhiqiang Wang. 2020. A review of research on object detection based on deep learning. In Journal of Physics: Conference Series, Vol. 1684. IOP Publishing, 012028.
Ross Girshick, Jeff Donahue, Trevor Darrell, and Jitendra Malik. 2015. Regionbased convolutional networks for accurate object detection and segmentation. IEEE transactions on pattern analysis and machine intelligence 38, 1 (2015), 142–158.
Jonas Gomes and Luiz Velho. 2002. Computação gráfica: imagem. IMPA.
Glenn Jocher. 2020. YOLOv5 by Ultralytics. https://doi.org/10.5281/zenodo.3908559
Bruno Henrique Pereira Marques. 2019. Avaliação de algoritmos baseados em deep learning para localizar placas veiculares brasileiras em ambientes complexos.
Ricardo Meier. 2021. Mapa de estações do Metrô e CPTM. [link]
Metrô-SP. 2013. Manual do Usuário com Deficiência. [link].
Metrô-SP. 2017. Portal da Transparência do Metrô - Pesquisa Origem e Destino. [link].
Metrô-SP. 2022. Portal da Transparência do Metrô - Demanda. [link].
Tobias Mettler, Stephan Daurer, Michael A. Bächle, and Andreas Judt. [n. d.]. Do-it-yourself as a means for making assistive technology accessible to elderly people: Evidence from the ICARE project. Information Systems Journal n/a, n/a ([n. d.]). https://doi.org/10.1111/isj.12352 arXiv: [link]
Amir Mukhtar. 2022. Vision based system for detecting and counting mobility aids in surveillance videos.
Amir Mukhtar, Michael J. Cree, Jonathan B. Scott, and Lee Streeter. 2018. Mobility Aids Detection Using Convolution Neural Network (CNN). In 2018 International Conference on Image and Vision Computing New Zealand (IVCNZ). 1–5. https://doi.org/10.1109/IVCNZ.2018.8634731
Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. 2016. You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition. 779–788.
Harichandana B S S, Vibhav Agarwal, Sourav Ghosh, Gopi Ramena, Sumit Kumar, and Barath Raj Kandur Raja. 2022. PrivPAS: A real time Privacy-Preserving AI System and applied ethics. In 2022 IEEE 16th International Conference on Semantic Computing (ICSC). 9–16. https://doi.org/10.1109/ICSC52841.2022.00010
Yuki Sakai, Huimin Lu, Joo-Kooi Tan, and Hyoungseop Kim. 2019. Recognition of surrounding environment from electric wheelchair videos based on modified YOLOv2. Future Generation Computer Systems 92 (2019), 157–161.
Konstantin Struebig, Niklas Ganter, Leon Freiberg, and Tim C Lueth. 2021. Stair and Ramp Recognition for Powered Lower Limb Exoskeletons. In 2021 IEEE International Conference on Robotics and Biomimetics (ROBIO). IEEE, 1270–1276.
Secretaria dos Direitos da Pessoa com Deficiência do Estado de São Paulo. 2021. Estimativa populacional de pessoas com deficiência. [link].
Marcelo Telles, Jorge Luis Barbosa, and Rodrigo Righi. 2016. Um Modelo Computacional para Acessibilidade em Cidades Inteligentes. In Anais do XII Simpósio Brasileiro de Sistemas de Informação (Florianópolis). SBC, Porto Alegre, RS, Brasil, 116–123. https://doi.org/10.5753/sbsi.2016.5953
Andres Vasquez, Marina Kollmitz, Andreas Eitel, and Wolfram Burgard. 2017. Deep detection of people and their mobility aids for a hospital robot. In 2017 European Conference on Mobile Robots (ECMR). IEEE, 1–7.
P Vijayalakshmi et al. 2021. Development of Speech and Gesture Enabled Wheelchair System for People with Cerebral Palsy. In 2021 3rd International Conference on Signal Processing and Communication (ICPSC). IEEE, 620–624.
Paul Viola and Michael Jones. 2001. Rapid object detection using a boosted cascade of simple features. In Proceedings of the 2001 IEEE computer society conference on computer vision and pattern recognition. CVPR 2001, Vol. 1. Ieee, I–I.