Theft or robbery of equipment at universities and research centers can cause financial losses, in the millions of reais, or even academic losses, which are invaluable, as they take with them long-standing research and work data, which are often difficult to recover. This is a more common reality than one might think: a quick search on the internet will reveal several reports of such incidents throughout the country.  

Currently, even with cameras installed on campus, the response time of security teams in these spaces, especially on weekends and holidays, is often slow and images may not be available in real time.  

Under development by the CampusEdge Working Group of the Advanced Services RD&I Program Phase 2 in 2024, the GuardAI solution can quickly identify cases of theft and/or theft of cables and equipment on university campuses and institutes, which allows for quick decision-making, improving the response time of campus security teams. 

GuardAI was created by a team from the Teleinformatics and Automation Group at the Federal University of Rio de Janeiro (GTA/UFRJ), who realized that using technology would significantly improve the perception of security on university campuses.  

The edge computing architecture adopted by the solution is a technical differential of the designed solution. This architecture allows real-time video processing, for detecting these events and generating alarms, to be performed at the edge, significantly increasing the solution's scalability potential, allowing this solution to be deployed in several organizations using the RNP System in the future.  

Through the video management systems in use in institutions, GuardAI, using artificial intelligence (AI) and computer vision techniques, enables the automated, real-time detection of suspicious events, promptly alerting security teams. “The idea behind GuardAI is not to replace security guards, but rather to direct their attention to the cameras that detect incidents. In addition, it is possible to record all detected incidents for later analysis,” says the WG coordinator, Rodrigo de Souza Couto. The solution uses video streams from cameras installed on campuses through open protocols. 

The edge processing modules can be configured in the cloud and initially allow the detection of people and objects. The ease of customization and the processing cost adaptable to the available budget are the main benefits of the platform. According to the WG coordinator, the solutions currently available are generic and require high hardware capacity.  

In addition to universities, healthcare establishments with teaching and research areas, museums and cultural institutions (which need to preserve their collections) can also benefit from GuardAI.  

Next steps: ONVIF integration and minimum viable product (MVP) validation  

The team is currently refining the object detection models, including people counting and pose estimation models, and incorporating anomalous behavior detection. The solution, currently under development, will be capable of using videos captured using commonly used industry protocols (e.g. RTSP and ONVIF).  

GuardAI is currently undergoing solution validation at UFRJ. If your institution would like to validate this solution on your campus, please send an email to: contato@guardai.tech. 

The solution will be presented in a webinar on 5/7, at 3 pm. Register here. 

Learn more about GuardAI here