CLNet-tracking

News

Our extension version (Adaptive Siamese Tracking with a Compact Latent Network) is accepted to T-PAMI. We obtain the raw results for comparison with other methods.

Content

Implementation code for

CLNet: A Compact Latent Network for Fast Adjusting Siamese Trackers. In Proceedings of the European Conference on Computer Vision (ECCV), 2020.

By Xingping Dong, Jianbing Shen, Ling Shao, Fatih Porikli.

========================================================================

Any comments, please email: xingping.dong@gmail.com, shenjianbingcg@gmail.com

This software was based on the PySOT and developed under Ubuntu 16.04 with python 3.7.

If you use this software for academic research, please consider to cite the following papers:

@article{dong2022adaptive,
  title={Adaptive Siamese Tracking with a Compact Latent Network},
  author={Dong, Xingping and Shen, Jianbing and Porikli, Fatih and Luo, Jiebo and Shao, Ling},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2022},
  publisher={IEEE}
}
@inproceedings{dong2020clnet,
  title={CLNet: A Compact Latent Network for Fast Adjusting Siamese Trackers},
  author={Dong, Xingping and Shen, Jianbing and Shao, Ling and Porikli, Fatih},
  booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
  pages={378--395},
  year={2020}
}
@article{dong2019dynamical,
  title={Dynamical Hyperparameter Optimization via Deep Reinforcement Learning in Tracking},
  author={Dong, Xingping and Shen, Jianbing and Wang, Wenguan and Shao, Ling and Ling, Haibin and Porikli, Fatih},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2019},
  publisher={IEEE}
}
@inproceedings{dong2018hyperparameter,
  title={Hyperparameter optimization for tracking with continuous deep q-learning},
  author={Dong, Xingping and Shen, Jianbing and Wang, Wenguan and Liu, Yu and Shao, Ling and Porikli, Fatih},
  booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition},
  pages={518--527},
  year={2018}
}
@article{dong2019quadruplet,
  title={Quadruplet network with one-shot learning for fast visual object tracking},
  author={Dong, Xingping and Shen, Jianbing and Wu, Dongming and Guo, Kan and Jin, Xiaogang and Porikli, Fatih},
  journal={IEEE Transactions on Image Processing},
  volume={28},
  number={7},
  pages={3516--3527},
  year={2019},
  publisher={IEEE}
}
@inproceedings{dong2018triplet,
  title={Triplet loss in siamese network for object tracking},
  author={Dong, Xingping and Shen, Jianbing},
  booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
  pages={459--474},
  year={2018}
}

Installation

Please find installation instructions for PyTorch and PySOT in INSTALL.md.

Pretrained Models and Raw results

We provide the original model (ResNet 50) in SiamRPN++ as the base model for training, and the pretrained model of CLNet for testing. Please download the above models and put them in the folder ./pretrained_models.

We also obtain the raw results for comparison with other methods.

Testing

Add PySOT to your PYTHONPATH

export PYTHONPATH=/path/to/pysot:$PYTHONPATH

Download models

Download the pretrained model of CLNet, and put it in the folder ./pretrained_models.

Webcam demo

python tools_leo/demo.py \
    --config experiments/clnet_r50_l234/config-vot.yaml \
    --snapshot pretrained_models/clnet_r50_l234.pth
    # --video demo/bag.avi # (in case you don't have webcam)

Download testing datasets

Download datasets and put them into testing_dataset directory. Jsons of OTB, VOT2016/2018/2019, NFS, UAV, LaSOT datasets can be downloaded from Google Drive or BaiduYun. Jsons of DTB, Temple-color-128 datasets can be downloaded from Google Drive. More details could be found in ./testing_dataset/README.md. If you want to test tracker on new dataset, please refer to pysot-toolkit to setting testing_dataset.

Test tracker

cd experiments/clnet_r50_l234
python -u ../../tools_leo/test_last_pos_neg2.py	\
	--snapshot ../../pretrained_models/clnet_r50_l234.pth 	\ # model path
	--dataset VOT2019 	\ # dataset name
	--config config-vot.yaml	  # config file

The testing results will in the current directory(results/dataset/model_name/)

Eval tracker

assume still in experiments/clnet_r50_l234

python ../../tools_leo/eval.py 	 \
	--tracker_path ./results \ # result path
	--dataset VOT2019        \ # dataset name
	--num 1 		 \ # number thread to eval
	--tracker_prefix 'cl*'   # tracker_name

Training

See TRAIN.md for detailed instruction.

Reproduction

We obtain the raw results for comparison with other methods.

To reproduce our results, we also provide the docker image in docker hub to run our codes. Please see experiments/clnet_r50_l234/README.md for more details.

xingpingdong/CLNet-tracking