SiamVGG: Visual Tracking with Deeper Siamese Networks
The accuracy may be improved by using cv2 instead of PIL. I'll developing it later (if you are interested in it you can modified it yourself).
Won 4th place in VOT2018 Realtime Challenge
SiamVGG with RPN module link
Required:
Python 2.7 PyTorch 4.0 (in VOTToolkit <=3.0)
- train: Including the training code.
- dataset: Instruction of generating the dataset note step by step.
- vot2018submission: Including the original submission in VOT2018. Download the pretrained data from Google Drive and follow the instruction to set up the test. (Partial of the code are modified from SiamFC-pytorch and [https://github.com/HelloRicky123/Siamese-RPN], thanks to Lianghua and Ruiqi.)
If you find our project useful, please cite our technical report:
@inproceedings{Li2019SiamVGGVT,
title={SiamVGG: Visual Tracking using Deeper Siamese Networks},
author={Yuhong Li and Xiaofan Zhang},
year={2019}
}