Python module bindings for SORT algorithm (Simple, Online, and Realtime Tracking) implemented in C++, with the goal of being fast and easy to use. Tracker is based on this repo.


Before you can install the package, you need to install the following dependencies:

$ sudo apt install libopencv-dev
$ sudo apt install libeigen3-dev

Make sure pip is upgraded to last version:

$ pip install pip --upgrade

Then you can install the package using:

$ pip install sort-tracker


$ git clone
$ cd sort-python
$ pip install .


import sort
# Create a tracker with max_coast_cycles = 5 and min_hits = 3
# Default values are max_coast_cycles = 3 and min_hits = 1
tracker = sort.SORT(5,3)


Two main methods are available named run and get_tracks and you can specify format of input and output bounding boxes as follows:

# format (int):
#    0: [xmin, ymin, w, h]
#    1: [xcenter, ycenter, w, h]
#    2: [xmin, ymin, xmax, ymax]

defaulf format is 0.


run method takes an array of bounding boxes and format, and then performs tracking.

# Input:
#   bounding_boxes: a numpy array of bounding boxes [n, 4]
#   format: format of bounding boxes (int)
import numpy as np
bounding_boxes = np.array([[10, 10, 20, 20], [30, 30, 40, 40]]), 0)


get_tracks method returns a array of tracks.

# Input:
#   format: format of bounding boxes (int)
# Output:
#   tracks: a numpy array of tracks [n, 5] where n is the number of tracks
#       and 5 is (id, ..., ..., ..., ...) where id is the track id and ... is the bounding box in the specified format
tracks = tracker.get_tracks(0)



Author: MrGolden1

Email: [email protected]


View Github