This is a pilot project developed by Urban Big Data Centre and Glasgow City Council with the aim of counting the number of persons and vehicles in cctv images.
To achieve this aim we make use of the tensorflow object detection API from Google, available in:
https://github.com/tensorflow/models/tree/master/research/object_detection
This API makes use of object detection algorithms to extract instances of real-world objects in images.
In this project a picture is taken every 15 minutes in a specific position - home position. Afterwards the images are processed with TensorFlow Object Detection API and the result of the count for each of the four cameras in the pilot is appended to a report csv file.
A server machine with OS Ubuntu Server was used to run the project. Please refer to the installation steps to know more about the configuration of the server.
Special thanks to Christina Boididou, a former employee of UBDC, who started the development of this project and who is responsible for some of the scripts and ideas presented here.