●
The ITRex team designed an AI-powered solution that runs on the edge of the network on standalone surveillance cameras. The cameras are enhanced with powerful chips that allow running trained AI algorithms locally.
●
The solution is powered by Beta-Variational Autoencoders (Beta-VAE) and Gaussian Mixture Models (GMM). To train the algorithms, the ITRex team collected around 3,000 hours of video footage showcasing normal behavior from the airport’s security cameras. During training, the algorithms learned to detect any deviations from the normal behavior and pinpoint those in new, unseen footage.
●
The intelligent surveillance cameras are integrated with the conveyor belt control module via RS-485 (think: a communication protocol used for transmitting data between different devices), which allows automatically stopping a conveyor belt once an anomaly is detected.