A general and large-scale benchmark COESOT dataset for color-event based visual tracking

---

Revisiting Color-Event based Tracking: A Unified Network, Dataset, and Metric, Chuanming Tang, Xiao Wang, Ju Huang, Bo Jiang, Lin Zhu, Jianlin Zhang, Yaowei Wang, Yonghong Tian [Project]

Update Log

• 🔥 [2024.03.12] A New Long-term RGB-Event based Visual Object Tracking Benchmark Dataset (termed FELT) is available at [Paper] [Code] [DemoVideo]

• 🔥 [2024.03.06] Tracking results of CEUTrack on VisEvent dataset is available at [ceutrack_visevent_dataset_tracking_results.zip]

• 🔥 [2023.09.27] A High Definition (HD) Event based Visual Object Tracking Benchmark Dataset (termed EventVOT) is available at [arXiv] [Github]

Demo Video:

• [YouTube]

Dataset Download:

Baidu Download link：https://pan.baidu.com/s/12XDlKABlz3lDkJJEDvsu9A     Passcode：AHUT 

The directory should have the below format:

├── COESOT dataset
    ├── Training Subset (827 videos, 160GB)
        ├── dvSave-2021_09_01_06_59_10
            ├── dvSave-2021_09_01_06_59_10_aps
            ├── dvSave-2021_09_01_06_59_10_dvs
            ├── dvSave-2021_09_01_06_59_10.aedat4
            ├── groundtruth.txt
            ├── absent.txt
            ├── start_end_index.txt
        ├── ... 
    ├── Testing Subset (528 videos, 105GB)
        ├── dvSave-2021_07_30_11_04_12
            ├── dvSave-2021_07_30_11_04_12_aps
            ├── dvSave-2021_07_30_11_04_12_dvs
            ├── dvSave-2021_07_30_11_04_12.aedat4
            ├── groundtruth.txt
            ├── absent.txt
            ├── start_end_index.txt
        ├── ... 

COESOT_eval_toolkit

1. unzip the COESOT_eval_toolkit.zip, and open it with Matlab (over Matlab R2020).

2. add your tracking results and baseline results (Passcode：siaw) in $/coesot_tracking_results/ and modify the name in $/utils/config_tracker.m. BTW, here we also provide the event-only baseline tracking methods results in [Event_only Results] Passcode：qblp

3. run Evaluate_COESOT_benchmark_SP_PR_only.m for the overall performance evaluation, including SR, PR, NPR.

4. run plot_BOC.m for BOC score evaluation and figure plot.
5. run plot_radar.m for attributes radar figrue plot.

6. run Evaluate_COESOT_benchmark_attributes.m for attributes analysis and figure saved in $/res_fig/.

CEUTrack

A unified framework for color-event tracking.

[Models] Passcode：0uk0 [Raw Results] Passcode：yeow [Training logs] Passcode：hnim

Install env

conda create -n event python=3.7
conda activate event
bash install.sh

Run the following command to set paths for this project

python tracking/create_default_local_file.py --workspace_dir . --data_dir ./data --save_dir ./output

After running this command, you can also modify paths by editing these two files

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

Then, put the tracking datasets COESOT in ./data.

Download pre-trained MAE ViT-Base weights and put it under $/pretrained_models

Download the model weights and put it on $/output/checkpoints/train/ceutrack

• [Note] More useful scripts can be found from:

https://github.com/Event-AHU/COESOT/tree/main/CEUTrack/scripts

Train & Test & Evaluation

    # train
    export CUDA_VISIBLE_DEVICES=0
    python tracking/train.py --script ceutrack --config ceutrack_coesot  \
    --save_dir ./output --mode multiple --nproc_per_node 1 --use_wandb  0
    # test
    python tracking/test.py   ceutrack ceutrack_coesot --dataset coesot --threads 4 --num_gpus 1
    # eval
    python tracking/analysis_results.py --dataset coesot  --parameter_name ceutrack_coesot

Test FLOPs, and Speed

Note: The speeds reported in our paper were tested on a single RTX 3090 GPU.

# Profiling ceutrack_coesot
python tracking/profile_model.py --script ceutrack --config ceutrack_coesot

Activation Visualization

Use the script from: [show_CAM.py]

from .show_CAM import getCAM
getCAM(response, curr_image, self.idx)

TODO List

• Paper (arXiv) release
• COESOT dataset release
• Evaluation Toolkit release
• Source Code release
• Tracking Models release

Acknowledgments

• Thanks for the OSTrack, PyTracking and ViT library for a quickly implement.

Citation:

@article{tang2022coesot,
  title={Revisiting Color-Event based Tracking: A Unified Network, Dataset, and Metric},
  author={Tang, Chuanming and Wang, Xiao and Huang, Ju and Jiang, Bo and Zhu, Lin and Zhang, Jianlin and Wang, Yaowei and Tian, Yonghong},
  journal={arXiv preprint arXiv:2211.11010},
  year={2022}
}

Star History