Skip to content

Latest commit

 

History

History
139 lines (112 loc) · 4.29 KB

README.md

File metadata and controls

139 lines (112 loc) · 4.29 KB

Exploiting Multimodal Spatial-temporal Patterns for Video Object Tracking [AAAI2025]

Official implementation of STTrack, including models and training&testing codes.

Models & Raw Results(Google Driver) Models & Raw Results(Baidu Driver:9527)

News

[Dec 30, 2024]

  • We release codes, models and raw results. Thanks for your star.

Introduction

  • A new unified multimodal spatial-temporal tracking framework (e.g. RGB-D, RGB-T, and RGB-E Tracking).

  • STTrack excels in multiple multimodal tracking tasks. We hope it will garner more attention for multimodal tracking.

Strong Performance

Tracker LasHeR RGBT234 VisEvent DepthTrack VOT22RGBD
STTrack 60.3 66.7 61.9 77.6 63.3

Usage

Installation

Create and activate a conda environment:

conda create -n STTrack python=3.8
conda activate STTrack

Install the required packages:

bash install_sttrack.sh

Data Preparation

Put the training datasets in ./data/. It should look like:

$<PATH_of_STTrack>
-- data
    -- DepthTrackTraining
        |-- adapter02_indoor
        |-- bag03_indoor
        |-- bag04_indoor
        ...
    -- LasHeR/train/trainingset
        |-- 1boygo
        |-- 1handsth
        ...
    -- VisEvent/train
        |-- 00142_tank_outdoor2
        |-- 00143_tank_outdoor2
        ...
        |-- trainlist.txt

Path Setting

Run the following command to set paths:

cd <PATH_of_STTrack>
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

Training

Dowmload the pretrained foundation model (OSTrack) and put it under ./pretrained/.

bash train.sh

You can train models with various modalities and variants by modifying train.sh.

Testing

For RGB-D benchmarks

[DepthTrack Test set & VOT22_RGBD]
These two benchmarks are evaluated using VOT-toolkit.
You need to put the DepthTrack test set to./Depthtrack_workspace/ and name it 'sequences'.
You need to download the corresponding test sequences at./vot22_RGBD_workspace/.

bash test_rgbd.sh

For RGB-T benchmarks

[LasHeR & RGBT234]
Modify the <DATASET_PATH> and <SAVE_PATH> in./RGBT_workspace/test_rgbt_mgpus.py, then run:

bash test_rgbt.sh

We refer you to LasHeR Toolkit for LasHeR evaluation, and refer you to MPR_MSR_Evaluation for RGBT234 evaluation.

For RGB-E benchmark

[VisEvent]
Modify the <DATASET_PATH> and <SAVE_PATH> in./RGBE_workspace/test_rgbe_mgpus.py, then run:

bash test_rgbe.sh

We refer you to VisEvent_SOT_Benchmark for evaluation.

Bixtex

If you find STTrack is helpful for your research, please consider citing:

@inproceedings{STTrack,
  title={Exploiting Multimodal Spatial-temporal Patterns for Video Object Tracking},
  author={Xiantao, Hu and Ying, Tai and Xu, Zhao and Chen, Zhao and Zhenyu, Zhang and Jun, Li and Bineng, Zhong and Jian, Yang},
  booktitle={AAAI},
  year={2025}
}

Acknowledgment

  • This repo is based on OSTrack and ViPT which are excellent works.
  • We thank for the PyTracking library, which helps us to quickly implement our ideas.