TrackMe

Simple Tracking Annotation Tool based on LabelMe

Description

This tool is built to integrate tracking visualization and annotation capabilities into LabelMe.
Please feel free to use if your project/work includes visualization of multi-object tracking and editting of object information.
TrackMe annotation format is compatible with LabelMe annotation format (.json) without conversion.

TrackMe saves and displays the tracking information of multiple objects on the right. It generates unique colors for different combinations of object label and ID

Features

• Add/remove tracking ID.
• Associate boxes (assign IDs) for existing non-ID detection boxes in the video folder (SORT).
• Interpolate boxes in long video range in case no pre-defined detection boxes.
• Modify/Delete boxes of same info throughout a list of continuous frames.
• Display homegeneous color for same object info (for the sake of multi-view object tracking).

Installation

Please install Anaconda environment on your computer beforehand.

conda create --name=trackGUI python=3.8
conda activate trackGUI
pip install -e .

Usage

conda activate trackGUI
labelme

Note

All frames must have labeled boxes

• Track from scratch: track from the first frame to the end frame with automatic ID assignment
• Track from Current Frame w/ Annotation: track from the current (being opened) frame with the modified ID or manual ID assignment
• Track from Current Frame w/0 Annotation: track from the current (being opened) frame with automatic ID assignment

Citation

If you find our work helpful, please consider citing our paper:

@article{phan2024trackme,
  title={TrackMe: A Simple and Effective Multiple Object Tracking Annotation Tool},
  author={Phan, Thinh and Phillips, Isaac and Lockett, Andrew and Kidd, Michael T and Le, Ngan},
  journal={arXiv preprint arXiv:2410.15518},
  year={2024}
}