• Tensorflow Version has been available by my classmates makalo. If you have any question, please feel free to contact us.
• This is a re-implementation for High Performance Visual Tracking with Siamese Region Proposal Network with PyTorch, which is accepted at CVPR2018.
• Code_v1.0 is available for traning, you should change your dataset as VOT format(top-left point and w,h). If there is a break in a sequence, ues "0,0,0,0" to replace the info of this frame.
• Dataset Tree

-root/class1/img1.jpg
            /...
            /imgN.jpg
            /groundtruth.txt

Paper: @InProceedings{Li_2018_CVPR,
author = {Li, Bo and Yan, Junjie and Wu, Wei and Zhu, Zheng and Hu, Xiaolin},
title = {High Performance Visual Tracking With Siamese Region Proposal Network},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2018}
}

• python=3.6
• pytorch=0.4.0
• cuda=9.0
• shapely=1.6.4

wget http://data.votchallenge.net/vot2013/vot2013.zip 

git clone https://github.com/mbuckler/youtube-bb.git
python3 download.py ./dataset 12

Pretrained model is available here BaiduYun

git clone https://github.com/songdejia/siamese-RPN
cd code_v1.0
python train_siamrpn.py --dataroot=/PATH/TO/YOUR/DATASET --lr=0.001 --checkpoint_path=/PATH/TO/YOUR/WEIGHT

bbox in detection
green -- ground truth which is got by pos anchor shift with reg_target
red -- bbox which is got by pos anchor with reg_pred
black -- bbox with highest score

proposal in original image

• Bo Li - paper - Siamese-RPN
• De jiasong - code - Siamese-RPN-pytorch
• Makalo - code - Siamese-RPN-tensorflow