Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit

* Fixed the display flag

* Update

Git stats


Failed to load latest commit information.
Latest commit message
Commit time
August 16, 2017 22:15
February 4, 2017 14:17
February 14, 2017 13:17
February 14, 2017 13:17
February 14, 2017 13:17



This repository contains code for Simple Online and Realtime Tracking with a Deep Association Metric (Deep SORT). We extend the original SORT algorithm to integrate appearance information based on a deep appearance descriptor. See the arXiv preprint for more information.


The code is compatible with Python 2.7 and 3. The following dependencies are needed to run the tracker:

  • NumPy
  • sklearn
  • OpenCV

Additionally, feature generation requires TensorFlow (>= 1.0).


First, clone the repository:

git clone

Then, download pre-generated detections and the CNN checkpoint file from here.

NOTE: The candidate object locations of our pre-generated detections are taken from the following paper:

F. Yu, W. Li, Q. Li, Y. Liu, X. Shi, J. Yan. POI: Multiple Object Tracking with
High Performance Detection and Appearance Feature. In BMTT, SenseTime Group
Limited, 2016.

We have replaced the appearance descriptor with a custom deep convolutional neural network (see below).

Running the tracker

The following example starts the tracker on one of the MOT16 benchmark sequences. We assume resources have been extracted to the repository root directory and the MOT16 benchmark data is in ./MOT16:

python \
    --sequence_dir=./MOT16/test/MOT16-06 \
    --detection_file=./resources/detections/MOT16_POI_test/MOT16-06.npy \
    --min_confidence=0.3 \
    --nn_budget=100 \

Check python -h for an overview of available options. There are also scripts in the repository to visualize results, generate videos, and evaluate the MOT challenge benchmark.

Generating detections

Beside the main tracking application, this repository contains a script to generate features for person re-identification, suitable to compare the visual appearance of pedestrian bounding boxes using cosine similarity. The following example generates these features from standard MOT challenge detections. Again, we assume resources have been extracted to the repository root directory and MOT16 data is in ./MOT16:

python tools/ \
    --model=resources/networks/mars-small128.pb \
    --mot_dir=./MOT16/train \

The model has been generated with TensorFlow 1.5. If you run into incompatibility, re-export the frozen inference graph to obtain a new mars-small128.pb that is compatible with your version:

python tools/

The stores for each sequence of the MOT16 dataset a separate binary file in NumPy native format. Each file contains an array of shape Nx138, where N is the number of detections in the corresponding MOT sequence. The first 10 columns of this array contain the raw MOT detection copied over from the input file. The remaining 128 columns store the appearance descriptor. The files generated by this command can be used as input for the

NOTE: If python tools/ raises a TensorFlow error, try passing an absolute path to the --model argument. This might help in some cases.

Training the model

To train the deep association metric model we used a novel cosine metric learning approach which is provided as a separate repository.

Highlevel overview of source files

In the top-level directory are executable scripts to execute, evaluate, and visualize the tracker. The main entry point is in This file runs the tracker on a MOTChallenge sequence.

In package deep_sort is the main tracking code:

  • Detection base class.
  • A Kalman filter implementation and concrete parametrization for image space filtering.
  • This module contains code for min cost matching and the matching cascade.
  • This module contains the IOU matching metric.
  • A module for a nearest neighbor matching metric.
  • The track class contains single-target track data such as Kalman state, number of hits, misses, hit streak, associated feature vectors, etc.
  • This is the multi-target tracker class.

The expects detections in a custom format, stored in .npy files. These can be computed from MOTChallenge detections using We also provide pre-generated detections.

Citing DeepSORT

If you find this repo useful in your research, please consider citing the following papers:

  title={Simple Online and Realtime Tracking with a Deep Association Metric},
  author={Wojke, Nicolai and Bewley, Alex and Paulus, Dietrich},
  booktitle={2017 IEEE International Conference on Image Processing (ICIP)},

  title={Deep Cosine Metric Learning for Person Re-identification},
  author={Wojke, Nicolai and Bewley, Alex},
  booktitle={2018 IEEE Winter Conference on Applications of Computer Vision (WACV)},


Simple Online Realtime Tracking with a Deep Association Metric







No releases published


No packages published