In this work, we evaluated the authenticity of drone in real time scenario using computer vision based deep tracker.

More information about this work is in this file

Python version 3.7 has been used.

Following installations are required for working code

Install the below libraries inside your conda environment

conda install pytorch torchvision -c pytorch

pip install opencv-python imutils pyyaml yacs tqdm colorama matplotlib cython tensorboardX

For using the code on other platforms except OSX, comment the following line from drone_tracking.py

os.environ['KMP_DUPLICATE_LIB_OK']='True'

Run the following code for running the program.

python drone_tracking.py --config experiments/siamrpn_alex_dwxcorr_otb/config.yaml --snapshot experiments/siamrpn_alex_dwxcorr_otb/model.pth

–config flag is for network configuration path

–snapshot flag is for network learned weights

Reference

Also, we utilized the code from pysot for SiamRPN.
For, more information on SiamRPN, kindly refer to the below work.

@INPROCEEDINGS{8579033,
  author={Li, Bo and Yan, Junjie and Wu, Wei and Zhu, Zheng and Hu, Xiaolin},
  booktitle={2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition}, 
  title={High Performance Visual Tracking with Siamese Region Proposal Network}, 
  year={2018},
  pages={8971-8980},
  doi={10.1109/CVPR.2018.00935}}

GitHub

View Github