Our Model Weight:
Pretrain Model: Google Drive(OSTrack)
Raw Result: Google Drive
💖 The implementation of MoE in this code is relatively simple. Please refer to my XTrack code, which contains a more detailed implementation. 💖
Create and activate a conda environment, we've tested on this env: You can follow the env setting of OSTrack.
Download the datasets from official project.
$<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
...
Run the following command to set paths:
cd <PATH_of_HotMoE>
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
- Download our model.
cd <PATH_of_HotMoE/tracking>
python test_hsi_mgpus_all.py --dataset_name HOT23VAL
If you want to test the real speed of our model, please run :
python test.py
@article{sun2025hotmoe,
title={HotMoE: Exploring Sparse Mixture-of-Experts for Hyperspectral Object Tracking},
author={Sun, Wenfang and Tan, Yuedong and Li, Jingyuan and Hou, Shuwei and Li, Xiaobo and Shao, Yingzhao and Wang, Zhe and Song, Beibei},
journal={IEEE Transactions on Multimedia},
year={2025},
publisher={IEEE}
}