- OpenCV 3.0+ (with Python binding)
- Python 2.7+, 3.4+, 3.5+
- NumPy 1.10+
- SciPy 0.16+
$ bash shells/download.sh
$ python scripts/convert_annotations.py
$ python scripts/convert_seqs.py
Each .seq
movie is separated into .png
images. Each image's filename is consisted of {set**}_{V***}_{frame_num}.png
. According to the official site, set06
~set10
are for test dataset, while the rest are for training dataset.
(Number of objects: 346621)
$ python tests/test_plot_annotations.py