ColorTrack

Human tracking based on color of clothes

ColorTrack is a module that can track humans based on the color of their upper body garment (e.g. shirt). It is developed in python3.6. ColorTrack uses detectron2 for human detection on a set of image data, kmeans clustering for shirt color extraction and uses a greedy algorithm for associating the detected humans of two consequent image frames.

Requirements

When the module was developed, the following where used:

python 3.6
opencv 4.2.0
cuda 9.2-10.2 with compatible torch, torchvision (find the compatible versions here)
numpy 1.19.2
sklearn 0.23.2
matplotlib 3.3.2
detectron2 pre-built (depending on the cuda and torch versions, for pre-built installation follow these instructions)

Instructions

In order to use this repository with a rosbag:

1. Follow the intructions How to export images from a bag file and save them to </your_path_to_the_exported_images>.
2. Open scripts/tracking.py and set </your_path_to_the_exported_images> in dirc parameter.