Anonymously recognizing people in public places using a thermal camera.
Thermal cameras will be used to measure the activities in public places. Recordings of public places are sensitive to privacy violations, and the images should be anonymous. Currently, measuring the activity in a public place is done manually. The goal of this project is to automate the manual effort, and to measure the activity more consistently.
How can we use thermal images to measure activity?
The main design challenge was recognizing activity from thermal imaging. The technique should be reliable and fast. We decided on using machine learning with hand annotated images to train a neural network. We have developed a system that accepts thermal images that are sent by a thermal camera and can process them into a dashboard containing the information derived from these images.