Spotlight on Hamed Tabkhi: Advancing Public Safety through AI/ML


Dr. Hamed Tabkhi, a professor at UNC Charlotte and keynote speaker for the 2024 Smart & Connected Communities PI Meeting, is at the forefront of revolutionizing public safety through innovative applications of artificial intelligence (AI) in the NSF’s Smart and Connected Communities (S&CC) program and beyond. 

Currently deployed in the city of Charlotte, his project “Building Safe and Secure Communities through Real-Time Edge Video Analytics” broadly leverages AI to address public safety concerns. While the project primarily focuses on the usage of computer vision, behavioral analysis and edge computing technologies, it also delves into ethical and privacy concerns of the community. 

The project first received funding in 2018 as a potential innovation to address growing tensions between the wider community of Charlotte and law enforcement, even as the need for more robust community security arose due to urbanization, a rise in crime, and a shortage of law enforcement. Using cutting edge technologies, Dr. Tabkhi's team has integrated testbeds for their research into a real-world setting, and are collaborating closely with the local community as the data comes in.

Dr. Tabkhi’s research team currently has two operational test beds with 80 cameras in total for testing, and two more test beds in negotiation stages. The test beds utilize existing CCTV cameras in the city and computer vision algorithms to maintain anonymity of the people caught on camera while maintaining the fidelity of each data capture. 

The hope is that utilizing AI in this way will shift the focus from appearance-based recognition tactics to behavioral analysis in the city of Charlotte to detect anomalies and predict potentially unsafe situations. The algorithms process visual data from human behaviors and from there identify potential threats.

To address the challenges of real-world deployment, the project incorporates edge computing. This enables the processing of data closer to the source, reducing latency and ensuring faster response times. Dr. Tabkhi's team has successfully developed algorithms optimized for edge devices, enhancing computational efficiency without compromising accuracy.

Recognizing the limitations of existing benchmarks, the project has pioneered the creation of novel data sets specifically tailored to address challenges encountered in real-world scenarios. The team has innovatively combined transformer models with graph isomorphism for tasks like animal detection, creating ensemble models that collaboratively enhance robustness against environmental noise.

A significant aspect of the project the team is currently focused on is the development of a user-friendly App (available on both iOS and Android)  that provides real-time insights generated by the AI algorithms. This app serves as a valuable tool for community members by providing them with visibility into the system's findings. 

The engagement of local communities in the project has correspondingly been instrumental in refining the technology to suit their specific safety needs. Meaning, how the technology is now implemented in Charlotte could look slightly different for a neighboring community.

Looking ahead, Dr. Tabkhi envisions the project making a significant impact on both safety and economic development in Charlotte and beyond. He hopes to scale the project to cover the entire city and extend its applications to address diverse safety concerns, including infrastructure safety and public transit. 

Dr. Tabkhi also highlighted the educational benefits of the project, noting that it has attracted local students to pursue graduate studies and conduct research at UNC Charlotte.


Dr. Tabkhi would like to give special thanks to his collaborators, Dr. Shannon Reid from the Department of Criminal Justice and Criminology at UNC Charlotte, and Ms. Jeri Guido, his community partner from Central Piedmont Community College.

Check out the related news features below:

WCNC Charlotte News:

PBS "Carolina Impact" Season 10, Episode 27 (minutes 9:20-15:20):

Submitted by Regan Williams on