Person Tracking

In this repository two ros packages are given

1. pubimage: package to publish images from webcam.
2. person_tracking : package to track people using pretrained tensorflow models and deepsort algorithm.

person_tracking Package

The object(person) detection is done using the pretrained Tensorflow models available at Tensorflow detection model zoo) trained on COCO dataset.

For tracking and recognition of the detected objects/person, deepsort app is used. Deepsort App uses a pretrained tf-model "mars-small128.pb" (link provided at their github page).

pubimage Package

Package to publish unaltered frames from webcam to rostopic camImage.

Using in Project

For using these packages, just copy then into src folder of ROS project, build the project using catkin_make and run the launch file. Two changes have to be made before using the packages.

1. Changing the python environment in following nodes (replace #!/home/ros-indigo/rospythonenv/bin/python with your python virtualenv)
   • person_tracking/src/person_tracking.py
   • person_tracking/src/tracking.py
   • pubimage/src/pubimage.py
2. Provide the path to the tensorflow models in launch file person_tracking/launch/tracker_launch.launch.