#ByteTrack训练自己数据集详细教程！！
Step1. Install ByteTrack.
git clone https://github.com/Double-zh/ByteTrack.git
cd ByteTrack
pip3 install -r requirements.txt
python3 setup.py developStep2. Install pycocotools.
pip3 install cython; pip3 install 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'Step3. Others
pip3 install cython_bboxdocker build -t bytetrack:latest .
# Startup sample
mkdir -p pretrained && \
mkdir -p YOLOX_outputs && \
xhost +local: && \
docker run --gpus all -it --rm \
-v $PWD/pretrained:/workspace/ByteTrack/pretrained \
-v $PWD/datasets:/workspace/ByteTrack/datasets \
-v $PWD/YOLOX_outputs:/workspace/ByteTrack/YOLOX_outputs \
-v /tmp/.X11-unix/:/tmp/.X11-unix:rw \
--device /dev/video0:/dev/video0:mwr \
--net=host \
-e XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
-e DISPLAY=$DISPLAY \
--privileged \
bytetrack:latest### 1. datasets
└——————VOCdevkit
| └——————VOC2012
| └——————Annotations
| └——————ImageSets
| └——————Main
| └——————JPEGImages
└—————— divide_dataset.py
The COCO pretrained YOLOX model can be downloaded from their [model zoo](https://github.com/Megvii-BaseDetection/YOLOX/tree/0.1.0). After downloading the pretrained models, you can put them under <ByteTrack_HOME>/pretrained.
三、准备模型配置文件{create a Exp file for your dataset && modify get_data_loader and get_eval_loader in your Exp file}
根据需求修改文件yolox_voc_s_ZZH.py的种类数，在路径"exps/example/custom/"文件夹下
class Exp(MyExp):
def __init__(self):
super(Exp, self).__init__()
self.num_classes = 2 #在这进行修改
self.depth = 0.33
self.width = 0.50
self.warmup_epochs = 1
Train with custom dataset
cd <ByteTrack_HOME>
python3 train.py -f exps/example/custom/yolox_voc_s_ZZH.py -d 1 -b 1 --fp16 -o -c pretrained/yolox_s.pthcd <ByteTrack_HOME>
python3 ZZH_track.py webcam -f exps/example/custom/yolox_voc_s_ZZH.py -c YOLOX_outputs/yolox_voc_s_ZZH/latest_ckpt.pth.tar --fp16 --fuse --save_result取消注释ZZH_track.py第227行代码，并注释第228行代码
```shell
cd <ByteTrack_HOME>
python3 ZZH_track.py video -f exps/example/custom/yolox_voc_s_ZZH.py -c YOLOX_outputs/yolox_voc_s_ZZH/latest_ckpt.pth.tar --fp16 --fuse --save_result
@article{zhang2021bytetrack,
title={ByteTrack: Multi-Object Tracking by Associating Every Detection Box},
author={Zhang, Yifu and Sun, Peize and Jiang, Yi and Yu, Dongdong and Yuan, Zehuan and Luo, Ping and Liu, Wenyu and Wang, Xinggang},
journal={arXiv preprint arXiv:2110.06864},
year={2021}
}
A large part of the code is borrowed from YOLOX, FairMOT, TransTrack and JDE-Cpp. Many thanks for their wonderful works.