Projects

Smart CCTV with Coral

Team project for Ambient AI Bootcamp & Competition in 2022.

Despite the abundance of CCTVs in South Korea, CCTVs do not play a direct role in immediately identifying danger. Rather, they are only sought after incidents to procure evidence. The main reason for this is the lack of employees who monitor CCTVs. A survey in 2021 found that an average CCTV control center worker in South Korea monitors 98 cameras alone.

Another problem situation we identified are the potential dangers in homes: namely household violence and falling accidents. Since they happen in homes it is hard to secure evidence or context information. Moreover, it is difficult for authorities such as the police and doctors who can provide help to identify early signs of danger due to privacy issues.

To address these two problems we propose an edge-based Smart CCTV that can detect violence and falling within 1 second from live camera data. We used Coral edge TPU to realize our solution.

Our solution can address dangers immediately, and reduce the workload of CCTV control center workers. Because camera data is not sent to an external server, it can be utilized with home cameras without privacy issues. CCTVs are run 24 hours and the total amount of data is huge, but data from each device is independent. Therefore CCTVs are an ideal setting for edge computation.

Presentation slides

Demo video

Including Everyone in Digital Healthcare

Team project for the Ericsson-LG Girls in ICT Hackathon: Bridge the Digital Divide in 2021.

Right after COVID-19, all offline exercise programs went down. Many young people turned to digital healthcare solutions such as workout apps and virtual exercise classes. However, the elderly who are less digitally literate rarely made use of such programs.

Previous research showed that exercise behavior of older people were highly influenced by their peers. To encourage seniors to exercise more actively, we created an IoT solution that links public exercise equipment all around the city. The solution provides a visualization of personal achievement and a local leaderboard in each park showing the total amount of exercise everyone who visited the park did for the past day/week/month. The leaderboard provides the participants a sense of community and can enhance communication among frequent users.

The solution was created with the Ericsson IoT kit and Node-Red.