Occupational Health and Safety, as well as risk prevention in industsry, have been thoroughly regulated with strict legislation in all developed countries. Successful prevention, in this context, is commonly supported by the systematic analysis of reports and observations that record deficiencies, dangerous conditions and critical events. Modern technologies, currelty fueling this analysis, include Internet of Things and Data Analytics, integrating their mechanisms in the vast majority of systems and platforms that monitor workplaces and train the appropriate safety models.
Admittedly, the most efficient "sensors" and intelligent "agents" capable of producing and maintaining a consistent stream of qualitatitve safety reports, are the custormers and the employees themselves. Introducting human-in-the-loop when monitoring and assessing safety measure and controls, can instantly upgrade the knowledge desnity of what is reported and how actionable it is by experts and auditors. The awareness and contribution of both clients and employees, can be trained and incentivized to provide the most relevant and contextually enriched information from the field.
Enable clients and employees to contribute reports from check points
Multi-modal analysis of reports fueling a health and safety knowledge model
Incetivize contribution and create awareness with rewards and credits
During the pandemic, the need for public awareness on health and safety was particularly highlighted. Health workers along with the public, have entered a longstanding struggle where competence, responsibility and monitoring, are essential for ensuring safety and protecting our way of life. Utilizing emerging technologies and investing on the most recent developments, Safety4ALL aims to aid this struggle and the serve the above cause. More specifically the project aims:
1. To motivate the public's involvement and contribuing, when safety risks are indentified in environments and venues they commonly use or work at.
2. To support experts (health & safety consultants, auditors, store managers) deal with reported risks and apply the appropriate measures as soon as possible.
Investing on existing know-how and extracting insights from a rich database of audits, Safety4ALL will offer its users access to a trained model and a currated registry of classified dangers, allowing the immediate report of relevant risks and empowering their prioritized handling by domain experts.
The 3rd Hellenic Conference on Occupational Health and Safety, was organized by the Hellenic Institute of Occupational Health and Safety, ELINYAE. It was held at the Athens Concert Hall, on November 10 and 11, 2022. The aim of the Conference was to record current health and safety challenges in the workplace in a human-centered perspective. With hundreds of professionals participating, the event offered a rich set of speeches and presentations, thematically focused on 21 distinct topics.
ERGONOMIA contributed with 5 speeches, 11 e-posters, while also being a sponsor of the conference. The main speech addressed "The use of AI technologies for health & safety risk assessment" and focused on the results of the Classify system, developed in the framework of the Safety4All project. At the project's booth, all supported technologies for creating Points of Interest (POIs) were installed and demonstrated, facilitating the piloting and evaluation of the platform's functions and services by many participants.
The 3rd Beyond 2023 International Exhibition "TECH4PEOPLE" took place on 24-26 May. More than 180 exhibitors from 36 countries participated. A separate booth was set up to showcase the full functionality of Safety4ALL, promoting the platform's features to potential users, such as health & safety managers or officers, local government bodies and the general public. In addition, the Safety4ALL mobile application was made available and was piloted by visitors, through the PlayStore & AppStore.
Karouzos, C., Paraskevopoulos, G., & Potamianos, A. (2021). UDALM: Unsupervised domain adaptation through language modeling. arXiv preprint arXiv:2104.07078. [link]
Ventoura, N., Palios, K., Vasilakis, Y., Paraskevopoulos, G., Katsamanis, N., & Katsouros, V. (2021, August). Theano: A Greek-speaking conversational agent for COVID-19. In Proceedings of the 1st Workshop on NLP for Positive Impact (pp. 36-46). [link]
Georgiou, E., Paraskevopoulos, G., & Potamianos, A. (2021). M3: MultiModal Masking Applied to Sentiment Analysis. In Interspeech (pp. 2876-2880). [link]
Paraskevopoulos, G., Georgiou, E., & Potamianos, A. (2022, May). Mmlatch: Bottom-up top-down fusion for multimodal sentiment analysis. In ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (pp. 4573-4577). IEEE. [link]
Georgiou, E., Kritsis, K., Paraskevopoulos, G., Katsamanis, A., Katsouros, V., & Potamianos, A. (2023, January). Regotron: Regularizing the Tacotron2 architecture via monotonic alignment loss. In 2022 IEEE Spoken Language Technology Workshop (SLT) (pp. 977-983). IEEE. [link]
Fraile, L. P., & Koulamas, C. (2022, June). Design and Evaluation of an Indoor Positioning System based on Mobile Devices. In 2022 11th Mediterranean Conference on Embedded Computing (MECO) (pp. 1-6). IEEE. [link]
Panagiotou, C., Fraile, L. P., & Koulamas, C. (2023, June). Detecting Health & Safety Hazards through AI and Edge Computing on Mobile Devices. In 2023 12th Mediterranean Conference on Embedded Computing (MECO) (pp. 1-4). IEEE. [link]
Paraskevopoulos, G., Pistofidis, P., Banoutsos, G., Georgiou, E., & Katsouros, V. (2022). Multimodal Classification of Safety-Report Observations. Applied Sciences, 12(12), 5781. [link]