In visual tracking dataset, a sequence of images has one object for each image, in contrast to object detection dataset that has multiple objects in each image. This annotation tool is suited for that.

Windows user can use an executable file. You can dowload from the Releases.

git clone https://github.com/daisatojp/labelTrack.git
cd labelTrack
pip install PyQt6==6.6.1 PyQt6-Qt6==6.6.3
# Windows PowerShell
# $Env:PYTHONPATH="."
python labelTrack

• Click Open Image in toolbar to open a folder containing a sequence of images.
• Click Open Label in toolbar to select label file to be saved.
• Click Create BBox in toolbar and make box by a mouse dragging in each image.
• Click Save in toolbar to save label file.

<x coord of top-left>,<y coord of top-left>,<width>,<height>
<x coord of top-left>,<y coord of top-left>,<width>,<height>
<x coord of top-left>,<y coord of top-left>,<width>,<height>
...
...
...

More concretely, see sample/label.txt.

• labelImg as a reference.