Add task category and paper link
#1
by nielsr HF Staff - opened
README.md
CHANGED
|
@@ -1,159 +1,172 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
---
|
| 9 |
-
|
| 10 |
-
##
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
| :
|
| 47 |
-
|
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
|
| 60 |
-
|:
|
| 61 |
-
|
|
| 62 |
-
|
|
| 63 |
-
|
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
*
|
| 76 |
-
|
| 77 |
-
|
|
| 78 |
-
|:
|
| 79 |
-
| **[
|
| 80 |
-
| **[
|
| 81 |
-
| **[
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
*
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
|
| 91 |
-
|
|
| 92 |
-
| **[
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
#
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
}
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
task_categories:
|
| 3 |
+
- object-detection
|
| 4 |
+
tags:
|
| 5 |
+
- multi-object-tracking
|
| 6 |
+
- MOT
|
| 7 |
+
- detection
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
## Beyond the Survey: A Systematic Empirical Study of Detection and Association in Visual MOT
|
| 11 |
+
|
| 12 |
+
Paper: [Beyond the Survey: A Systematic Empirical Study of Detection and Association in Visual MOT](https://huggingface.co/papers/2609.22291)
|
| 13 |
+
|
| 14 |
+
Code: https://github.com/linh-gist/VisualMOT
|
| 15 |
+
|
| 16 |
+
Official implementation repository for the paper **"Beyond the Survey: A Systematic Empirical Study of Detection and Association in Visual MOT"**, accepted for publication in *Artificial Intelligence Review*.
|
| 17 |
+
|
| 18 |
+
> **Note:** This repository also includes the official Python implementation for:
|
| 19 |
+
> **"Adaptive Confidence Threshold for ByteTrack in Multi-Object Tracking"** (*ICCAIS 2023*, [arXiv 2312.01650](https://arxiv.org/abs/2312.01650)) at this folder [bytetrack](trackers/bytetrack).
|
| 20 |
+
|
| 21 |
+
---
|
| 22 |
+
|
| 23 |
+
### Overview
|
| 24 |
+
|
| 25 |
+
This repository provides tracking implementations for algorithms evaluated in our study: **SORT, DeepSORT, MOTDT, FairMOT, ByteTrack (AdaptByteTrack), OCSORT, DeepOCSORT, and VisualRFS**. We provide CMC (Camera Motion Compensation) for ByteTrack (AdaptByteTrack), VisualRFS, and OC-SORT. Evaluation Scores/Metrics are:
|
| 26 |
+
- [CLEAR](https://link.springer.com/article/10.1155/2008/246309), [IDF1](https://arxiv.org/pdf/1609.01775), [HOTA](https://arxiv.org/abs/2009.07736) evaluation available at [JonathonLuiten/TrackEval](https://github.com/JonathonLuiten/TrackEval).
|
| 27 |
+
- Tracking Effort Measure [TEM](https://arxiv.org/abs/2212.08536) ([vpulab/MOT-evaluation](https://github.com/vpulab/MOT-evaluation)).
|
| 28 |
+
|
| 29 |
+
For other evaluated methods, please clone their respective repositories and follow the authors' original execution instructions.
|
| 30 |
+
|
| 31 |
+
---
|
| 32 |
+
|
| 33 |
+
### Datasets & Detection Outputs
|
| 34 |
+
|
| 35 |
+
#### Pre-extracted Detections can be downloaded from Hugging Face ([linhmv/VisualMOT](https://huggingface.co/datasets/linhmv/VisualMOT)).
|
| 36 |
+
We provide extracted detection files with confidence scores $[0, 1]$ in the ([`./dets/`](./dets/)) directory:
|
| 37 |
+
|
| 38 |
+
| Detector | Venue / Source | Paper / Link |
|
| 39 |
+
|:-------------------------------| :--- | :--- |
|
| 40 |
+
| POI: `detector_poi` | ECCV 2016 | [arXiv:1610.06136](https://arxiv.org/pdf/1610.06136) |
|
| 41 |
+
| JDE: `detector_jde` | ECCV 2020 | [arXiv:1909.12605](https://arxiv.org/abs/1909.12605) |
|
| 42 |
+
| TraDeS: `detector_trades` | CVPR 2021 | [arXiv:2103.08808](https://arxiv.org/abs/2103.08808) |
|
| 43 |
+
| FairMOT: `detector_fairmot128` | IJCV 2021 | [arXiv:2004.01888](https://arxiv.org/abs/2004.01888) |
|
| 44 |
+
| GSDT: `detector_gsdt` | ICRA 2021 | [arXiv:2006.13164](https://arxiv.org/abs/2006.13164) |
|
| 45 |
+
| CSTrack: `detector_cstrack` | TIP 2022 | [arXiv:2010.12138](https://arxiv.org/abs/2010.12138) |
|
| 46 |
+
| YOLOX: `detector_bytetrack` | ECCV 2022 | [arXiv:2110.06864](https://arxiv.org/abs/2110.06864) |
|
| 47 |
+
| YOLOv11: `detectors_yolov11` | arXiv 2024 | [arXiv:2410.17725](https://arxiv.org/abs/2410.17725) |
|
| 48 |
+
|
| 49 |
+
- POI: Relies on [ETHZ](https://ieeexplore.ieee.org/document/4587581), [Caltech Pedestrian](https://ieeexplore.ieee.org/abstract/document/5206631), and a self-collected surveillance dataset.
|
| 50 |
+
- TraDeS: Utilizes a [CrowdHuman](https://arxiv.org/abs/1805.00123) pre-trained model for 2D tracking alongside the [MOTChallenge](https://motchallenge.net/) dataset.
|
| 51 |
+
- JDE, FairMOT, CSTrack, GSDT: Fine-tuned on the "[Mix of Six](https://github.com/Zhongdao/Towards-Realtime-MOT/blob/master/DATASET_ZOO.md)" dataset, which combines [Caltech Pedestrian](https://ieeexplore.ieee.org/abstract/document/5206631), [CityPersons](https://arxiv.org/abs/1702.05693), [ETHZ](https://ieeexplore.ieee.org/document/4587581), [MOTChallenge](https://motchallenge.net/), [CUHK-SYSU](https://arxiv.org/abs/1604.01850), and [PRW](https://arxiv.org/abs/1604.02531).
|
| 52 |
+
- YOLOX: Trained on a combination of [MOTChallenge](https://motchallenge.net/), [CrowdHuman](https://arxiv.org/abs/1805.00123), [CityPersons](https://arxiv.org/abs/1702.05693), and [ETHZ](https://ieeexplore.ieee.org/document/4587581).
|
| 53 |
+
- YOLOv11: `YOLO11x`, Only trained on [COCO Detection](https://cocodataset.org/#overview) Dataset
|
| 54 |
+
|
| 55 |
+
|
| 56 |
+
|
| 57 |
+
#### Evaluation Datasets
|
| 58 |
+
Dataset Name | Year | Source|
|
| 59 |
+
| :--- |:---------------------------------------------------------------------------------------------| :--- |
|
| 60 |
+
| [MOTChallenge](https://motchallenge.net/)| [2016](https://arxiv.org/abs/1603.00831), [2017](https://motchallenge.net/data/MOT17/), [2020](https://arxiv.org/abs/2003.09003) |
|
| 61 |
+
|[DanceTrack](https://arxiv.org/abs/2111.14690)| CVPR 2022 |[DanceTrack/DanceTrack](https://github.com/DanceTrack/DanceTrack)
|
| 62 |
+
|[SportsMOT](https://arxiv.org/abs/2304.05170)| ICCV 2023 |[MCG-NJU/SportsMOT](https://github.com/MCG-NJU/SportsMOT)
|
| 63 |
+
|[CrowdTrack](https://arxiv.org/abs/2507.02479)| arXiv 2025 | [loseevaya/CrowdTrack](https://github.com/loseevaya/CrowdTrack)
|
| 64 |
+
|
| 65 |
+
---
|
| 66 |
+
|
| 67 |
+
### Evaluated Tracking Algorithms
|
| 68 |
+
|
| 69 |
+
#### 1. Analytical Data Association
|
| 70 |
+
*Hand-crafted motion & appearance models*
|
| 71 |
+
|
| 72 |
+
| Method | Ref. Index | Year | Source Code |
|
| 73 |
+
|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| :---: |:------------:| :--- |
|
| 74 |
+
| **[SORT](https://arxiv.org/abs/1602.00763)** | [23] | ICIP 2016 | [abewley/sort](https://github.com/abewley/sort) |
|
| 75 |
+
| **[DeepSORT](https://arxiv.org/abs/1703.07402)** | [27] | ICIP 2017 | [nwojke/deep_sort](https://github.com/nwojke/deep_sort) |
|
| 76 |
+
| **[MOTDT](https://arxiv.org/abs/1809.04427)** | [28] | ICME 2018 | [longcw/MOTDT](https://github.com/longcw/MOTDT) |
|
| 77 |
+
| **[FairMOT](https://arxiv.org/abs/2004.01888)** | [29] | IJCV 2021 | [ifzhang/FairMOT](https://github.com/ifzhang/FairMOT) |
|
| 78 |
+
| **[ByteTrack](https://arxiv.org/abs/2110.06864)** | [24] | ECCV 2022 | [FoundationVision/ByteTrack](https://github.com/FoundationVision/ByteTrack) |
|
| 79 |
+
| **[AdaptByteTrack](https://arxiv.org/abs/2312.01650)** | [75] | ICCAIS 2023 | [linh-gist/AdaptConfByteTrack](https://github.com/linh-gist/AdaptConfByteTrack) |
|
| 80 |
+
| **[OCSORT](https://arxiv.org/abs/2203.14360)** | [25] | CVPR 2023 | [noahcao/OC_SORT](https://github.com/noahcao/OC_SORT) |
|
| 81 |
+
| **[DeepOCSORT](https://arxiv.org/abs/2302.11813)** | [30] | ICIP 2023 | [gerardmaggiolino/deep-oc-sort](https://github.com/gerardmaggiolino/deep-oc-sort) |
|
| 82 |
+
| **[StrongSORT](https://arxiv.org/abs/2202.13514)** | [32] | TMM 2023 | [dyhBUPT/StrongSORT](https://github.com/dyhBUPT/StrongSORT) |
|
| 83 |
+
| **[VisualRFS](https://arxiv.org/abs/2407.08872)** | [1] | PR 2024 | [linh-gist/VisualRFS](https://github.com/linh-gist/VisualRFS) |
|
| 84 |
+
| **[HybridSORT](https://arxiv.org/abs/2308.00783)** | [31] | AAAI 2024 | [ymzis69/HybridSORT](https://github.com/ymzis69/HybridSORT) |
|
| 85 |
+
| **[TrackTrack](https://openaccess.thecvf.com/content/CVPR2025/html/Shim_Focusing_on_Tracks_for_Online_Multi-Object_Tracking_CVPR_2025_paper.html)** | [26] | CVPR 2025 | [kamkyu94/TrackTrack](https://github.com/kamkyu94/TrackTrack) |
|
| 86 |
+
|
| 87 |
+
#### 2. Deep Learning Data Association
|
| 88 |
+
*Learned feature-based association*
|
| 89 |
+
|
| 90 |
+
| Method | Ref. Index | Year | Source Code |
|
| 91 |
+
|:-----------------------------------------------------------------------------------------------------| :---: |:-----------:| :--- |
|
| 92 |
+
| **[SUSHI](https://arxiv.org/abs/2212.03038)** | [33] | CVPR 2023 | [dvl-tum/sushi](https://github.com/dvl-tum/sushi) |
|
| 93 |
+
| **[LTTrack](https://ieeexplore.ieee.org/abstract/document/10536914)** | [34] | TCSVT 2024 | [linjiaping1/LTTrack](https://github.com/linjiaping1/LTTrack) |
|
| 94 |
+
| **[LG-MOT](https://arxiv.org/abs/2406.04844)** | [35] | TCSVT 2025 | [weslee88524/lg-mot](https://github.com/weslee88524/lg-mot) |
|
| 95 |
+
|
| 96 |
+
#### 3. End-to-End (E2E) Data Association
|
| 97 |
+
*Joint detection and association learning*
|
| 98 |
+
|
| 99 |
+
| Method | Ref. Index | Year | Source Code |
|
| 100 |
+
|:------------------------------------------------------------------------------| :---: |:----------:| :--- |
|
| 101 |
+
| **[MOTR](https://arxiv.org/pdf/2105.03247)** | [17] | ECCV 2022 | [megvii-research/MOTR](https://github.com/megvii-research/MOTR) |
|
| 102 |
+
| **[MeMOTR](https://arxiv.org/abs/2307.15700)** | [41] | ICCV 2023 | [mcg-nju/memotr](https://github.com/mcg-nju/memotr) |
|
| 103 |
+
| **[MOTIP](https://arxiv.org/abs/2403.16848)** | [42] | CVPR 2025 | [MCG-NJU/MOTIP](https://github.com/MCG-NJU/MOTIP) |
|
| 104 |
+
| **[CO-MOT](https://arxiv.org/abs/2305.12724)** | [43] | ICLR 2025 | [BingfengYan/CO-MOT](https://github.com/BingfengYan/CO-MOT) |
|
| 105 |
+
| **[SambaMOTR](https://arxiv.org/abs/2410.01806)** | [39] | ICLR 2025 | [mattiasegu/sambamotr](https://github.com/mattiasegu/sambamotr) |
|
| 106 |
+
|
| 107 |
+
|
| 108 |
+
|
| 109 |
+
|
| 110 |
+
|
| 111 |
+
### Usage
|
| 112 |
+
1. **Set Up Python Environment**
|
| 113 |
+
- Create a `conda` Python environment and activate it:
|
| 114 |
+
```sh
|
| 115 |
+
conda create --name virtualenv python==3.8.0
|
| 116 |
+
conda activate virtualenv
|
| 117 |
+
```
|
| 118 |
+
- lone this repository recursively to have pybind11
|
| 119 |
+
```sh
|
| 120 |
+
git clone --recursive https://github.com/linh-gist/AdaptConfByteTrack.git
|
| 121 |
+
```
|
| 122 |
+
- Install Packages
|
| 123 |
+
```sh
|
| 124 |
+
numpy==1.23.1
|
| 125 |
+
opencv-python==4.9.0.80
|
| 126 |
+
loguru==0.7.2
|
| 127 |
+
scipy==1.10.1
|
| 128 |
+
lap==0.5.12
|
| 129 |
+
cython_bbox==0.1.5
|
| 130 |
+
matplotlib==3.5.3
|
| 131 |
+
filterpy==1.4.5
|
| 132 |
+
motmetrics==1.4.0
|
| 133 |
+
openpyxl==3.1.5
|
| 134 |
+
pycocotools==2.0.7
|
| 135 |
+
tabulate==0.9.0
|
| 136 |
+
# git clone https://github.com/JonathonLuiten/TrackEval.git
|
| 137 |
+
# cd TrackEval, python setup.py build develop
|
| 138 |
+
```
|
| 139 |
+
|
| 140 |
+
2. **Prepare Data**
|
| 141 |
+
- Datasets:
|
| 142 |
+
- MOT16, MOT17, MOT20, DanceTrack, SportsMOT, CrowdTrack
|
| 143 |
+
- You can also run with your custom dataset but need a detector
|
| 144 |
+
|
| 145 |
+
3. **Run the Tracking Demo**
|
| 146 |
+
- Change parameters in `make_parser()` in `track.py` such as `use_gmc`, `data_dir` (MOTChallenge GT data)
|
| 147 |
+
- Run `python track.py`
|
| 148 |
+
|
| 149 |
+
|
| 150 |
+
### Citation
|
| 151 |
+
If you find this project useful in your research, please consider citing by:
|
| 152 |
+
|
| 153 |
+
```
|
| 154 |
+
@article{van2026beyond,
|
| 155 |
+
title={Beyond the Survey: A Systematic Empirical Study of Detection and Association in Visual MOT},
|
| 156 |
+
author={Linh Van Ma and Juhua Hu and Wei Cheng and Unse Fatima and Moongu Jeon},
|
| 157 |
+
booktitle={Artificial Intelligence Review},
|
| 158 |
+
year={2026},
|
| 159 |
+
publisher={Springer}
|
| 160 |
+
}
|
| 161 |
+
@inproceedings{van2023adaptive,
|
| 162 |
+
title={Adaptive Confidence Threshold for ByteTrack in Multi-Object Tracking},
|
| 163 |
+
author={Linh Van Ma and Muhammad Ishfaq Hussain and JongHyun Park and Jeongbae Kim and Moongu Jeon},
|
| 164 |
+
booktitle={2023 12th International Conference on Control, Automation and Information Sciences (ICCAIS)},
|
| 165 |
+
pages={370--374},
|
| 166 |
+
year={2023},
|
| 167 |
+
organization={IEEE}
|
| 168 |
+
}
|
| 169 |
+
```
|
| 170 |
+
|
| 171 |
+
### Acknowledgement
|
| 172 |
+
A part of the code is borrowed from [SORT](https://github.com/abewley/sort), [DeepSORT](https://github.com/nwojke/deep_sort), [MOTDT](https://arxiv.org/abs/1809.04427), [FairMOT](https://arxiv.org/abs/2004.01888), [ByteTrack](https://arxiv.org/abs/2110.06864), [OCSORT](https://arxiv.org/abs/2203.14360), [DeepOCSORT](https://arxiv.org/abs/2302.11813), and [VisualRFS](https://arxiv.org/abs/2407.08872). Thanks for their wonderful works.
|