Robust Tracking Module

Implementation of the Robust Tracking Module on SiamRPN

Pretrained model

GDrive

Setup environment

conda env create -f environment.yaml
conda activate rtm

Training

coming soon

Run tracking

Dataset evaluation

• Set correct paths for each dataset in configs/local.py
• Download pretrained model and save in checkpoints/model_RTM.pth
• python tracking_evaluation.py --dataset lasot --distortion original

Run demo

• Run on video: python demo.py path/to/video.mp4
• Run on webcam: python demo.py 0
• Distort input with --distortion argument:
  • White Gaussian Noise: WGN
  • Salt and Pepper: SnP
  • Gaussian Blur: GB
• Select initial bounding box with mouse (click and drag)

Citation

@article{karakostas2025enhancing,
  title={Enhancing visual object tracking robustness through a lightweight denoising module},
  author={Karakostas, Iason and Mygdalis, Vasileios and Nikolaidis, Nikos and Pitas, Ioannis},
  journal={The Visual Computer},
  pages={1--18},
  year={2025},
  publisher={Springer}
}

Related repositories

• DaSiamRPN
• Siamese-RPN-pytorch
• SiamRPN Pytorch
• GOT-10k toolkit

Acknowledgments

This work has received funding from the research project ”Energy Efficient and Trustworthy Deep Learning - DeepLET” is implemented in the framework of H.F.R.I call “Basic research Financing (Horizontal support of all Sciences)” under the National Recovery and Resilience Plan “Greece 2.0” funded by the European Union – NextGenerationEU (H.F.R.I. Project Number: 16762). This publication reflects the authors’ views only. The European Commission is not responsible for any use that may be made of the information it contains.