This is the official Python evaluation toolkit for VastTrack: Vast Category Visual Object Tracking.
[Mar. 28, 2024]
- Python evaluation toolkit for VastTrack is now available, which you can use to conveniently test your tracker on VastTrack.
First, make sure you have downloaded our VastTrack dataset and the test set list we provided. Then, as shown below, modify line 7 in dataset.py to the path of the VastTrack dataset and the path of the test set list.
dataset_dict = dict(
vasttrack=DatasetInfo(module="vasttrack", class_name="VastTrackDataset", kwargs=dict(base_path="/path/to/vasttrack", txt_file_path="/path/to/vasttrack_test_set_list.txt"))
)
Format the trackers' results. They should look like this:
-- trackers_results
-- SeqTrack
|-- Aardwolf-1.txt
|-- Aardwolf-2.txt
...
-- OSTrack
|-- Aardwolf-1.txt
|-- Aardwolf-2.txt
...
-- STARK
|-- Aardwolf-1.txt
|-- Aardwolf-2.txt
...
-- ARTrack
|-- Aardwolf-1.txt
|-- Aardwolf-2.txt
...
...
Our toolkit supports two functions:
- Simply print the trackers' Success rate, Precision, Normalized Precision on VastTrack.
python analysis_results.py --results_dir /path/to/trackers/results --to_print_results True
- Genarate plots of trackers' Success rate, Precision, Normalized Precision on VastTrack.
python analysis_results.py --results_dir /path/to/trackers/results --to_plot_results True --save_result_plot True --save_result_plot_dir ./
- Thanks for the PyTracking library, which helps us to build our toolkit.