DMTrack

The official implementation for the ICRA 2026 paper DMTrack: Spatio-Temporal Multimodal Tracking via Dual-Adapter.

Models

Models & Raw Results (Google Drive)

Usage

Data Preparation

Download the training datasets and put the training datasets in ./data/, It should look like:

$<PATH_of_DMTrack>
-- 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_DMTrack>
python tracking/create_default_local_file.py --workspace_dir . --data_dir <PATH_of_Datasets> --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) (Baidu Driver: 4lvo) / foundation model (Google Drive) and put it under ./pretrained/.

bash train_dmtrack.sh

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

We refer you to use LasHeR Toolkit for LasHeR evaluation, and refer you to use 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 eval_rgbe.sh

We refer you to use VisEvent_SOT_Benchmark for evaluation.

Citation

Please cite our work if you think it is useful for your research.

@inproceedings{DMTrack,
  title={DMTrack: Spatio-Temporal Multimodal Tracking via Dual-Adapter},
  author={Li, Weihong and Dong, Shaohua and Lu, Haonan and Zhang, Yanhao and Fan, Heng and Zhang, Libo},
  booktitle={2026 IEEE International Conference on Robotics and Automation (ICRA)},
  year={2025}
}

Acknowledgment

• This repo is based on ViPT and BAT, these exellent works help us to quickly implement our ideas.
• Thanks for the OSTrack and PyTracking library.