README.md

FairMOTVehicle

A fork of FairMOT used to do vehicle MOT(multi-object tracking). You can refer to origin fork
FairMOT link

车辆跟踪，效果如下，此测试未经过训练(Results of vehicle mot is as follows, the video seq has not been trained)：

使用UA-DETRAC公开数据集训练FairMOT(Using UA-DETRAC as training dataset for vehicle tracking)

UA_DETRAC是一个公开的车辆跟踪数据集, 共8万多张训练数据集，每一张图的每一辆车都经过了精心的标注。
UA-DETRAC benchmark link

训练方法(具体调用时，根据服务器目录, 修改自定义路径)

(1). 使用gen_labels_detrac.py脚本预处理原始的训练数据(Call gen_labels_detrac.py to prepare UA-DETRAC for training)

• 调用preprocess函数创建用于FairMOT的标准训练数据目录(Call preprocess function to make directory structure FairMOT required)
• 调用核心函数, gen_labels函数，解析UA-DETRAC的xml格式标签文件转换成FairMOT格式的标签文件，生成txt标签文件(Call the core function gen_labels to parse xml labels of UA-DETRAC and convert to the FairMOT label format, i.e txt file for each image)
• 调用gen_dot_train_file函数，生成用于训练的.train文件(Call gen_dot_train_file to generate dot train file for training(which contain all image path of training dataset))

(2). 编写json格式的cfg文件./src/lib/cfg/detrac.json(Edit a json configuration file for this training)

(3). 修改opts.py文件，修改训练参数，开始训练(Edit opts.py for trainig parameters)

• 修改--load_model参数, 选择一个断点模型, 如 ctdet_coco_dla_2x.pth, 从这个预训练模型开始训练
• 修改----data_cfg参数, 选择训练、测试数据，如 ../src/lib/cfg/detrac.json
• python or python3 ./src/train.py启动训练进程即可。