Vision-Based Multiobject Tracking Through UAV Swarm
Hao Shen, Defu Lin, Xiwen Yang, Shaoming He
The code of our paper Vision-Based Multiobject Tracking Through UAV Swarm. The code is based on the VehicleNet. The video of the constructed situation in AirSim simulation system is available at YouTube.
you should first download the MUMO dataset.
you should run object detection on MUMO dataset. The results format is
x_left_top, y_left_top, x_right_bottom, y_right_bottom
We open our detection results in the compressed file "detection.zip". For fair comparison, we suggest you use the same the directory detection results for different solutions of multi-UAV multi-object tracking. If you run other object detectors, the directory organization of results should be same as ours.
You should revise these directory paths at lines 116-118 in the file "main.py".
path_dataset = '/' #the directoy path of MUMO dataset
path_det = 'result/detections/' #the directoy path of object detections
path_out = './result/id/' #the results of the solution.