ViPT_HOT2023

ViPT variants for Hyperspectral Object Tracking Challenge 2023 (https://www.hsitracking.com/)

See our ViPT CVPR2023 Paper and PyTorch Code here!

🏆 Without bells and whistles, we won the third place using single model of ViPT variant.

Models & Raw Results (Google Driver) Video Presentation(YouTube)

Usage

Installation

Create and activate a conda environment, we've tested on this env:

timm                    0.6.11
torch                   1.9.1+cu102
torchfile               0.1.0
torchvision             0.10.1+cu102
python                  3.7.10

Data Preparation

Download the validation and ranking datasets from official project. It should look like:

$<PATH_of_Data>
-- validation
    -- HSI-NIR
        |-- basketball3
        ...
    -- HSI-NIR-FalseColor
        |-- basketball3
        ...
    -- HSI-RedNIR
        |-- ball&mirror9
        ...
    -- HSI-RedNIR-FalseColor
        |-- ball&mirror9
        ...
    -- HSI-VIS
        |-- ball
        ...
    -- HSI-VIS-FalseColor
        |-- ball
        ...
-- ranking
    -- HSI-NIR
        |-- basketball2
        ...
    -- HSI-NIR-FalseColor
        |-- basketball2
        ...
    -- HSI-RedNIR
        |-- backpack4
        ...
    -- HSI-RedNIR-FalseColor
        |-- backpack4
        ...
    -- HSI-VIS
        |-- cards6
        ...
    -- HSI-VIS-FalseColor
        |-- cards6
        ...

Path Setting

Run the following command to set paths:

cd <PATH_of_ViPT_HOT2023>
python tracking/create_default_local_file.py --workspace_dir . --data_dir ./data --save_dir ./output

You can also modify paths by these two files:

./lib/train/admin/local.py  # paths for training
./lib/test/evaluation/local.py  # paths for testing

Testing

Download pretrained model ViPT_all.pth.tar from link and put it into dir final_model.
Modify the line177 and line184 of the file tracking/test_hsi_mgpus_all.py to yours.

cd <PATH_of_ViPT_HOT2023/tracking>
python test_hsi_mgpus_all.py --dataset_name HOT23VAL
python test_hsi_mgpus_all.py --dataset_name HOT23TEST

Citation

If you find ViPT is helpful and use this code for your research, please consider citing:

@inproceedings{zhu2023visual,
  title={Visual prompt multi-modal tracking},
  author={Zhu, Jiawen and Lai, Simiao and Chen, Xin and Wang, Dong and Lu, Huchuan},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={9516--9526},
  year={2023}
}

laisimiao/ViPT_HOT2023