JDE

KO | EN

gitlite — search

#ai #typescript #deep-learning #docker #macos #windows #cli #kubernetes #awesome-list #vue #pytorch #ios #openai #open-source

(★ 29)

Towards Real-Time Multi-Object Tracking

JDE 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# JDE
Towards Real-Time Multi-Object Tracking

The code in the mot directory is easier to use when training. Use tools/train.py instead of ./train.py. The *.py file in the root directory will be deprecated.

<del>以下操作步骤均以MOT16为例</del>

<del>## 1.1 准备数据集</del>

<del>* 从MOT挑战赛官网下载数据集并解压 </del>
<del>wget https://motchallenge.net/data/MOT16.zip -P /data/tseng/dataset/jde </del>
<del>cd /data/tseng/dataset/jde </del>
<del>unzip MOT16.zip -d MOT16 </del>

<del>* 创建MOT16任务的工作区, 并将MOT格式标注文件转换为需要格式的标注文件 </del>
<del>git clone https://github.com/CnybTseng/JDE.git </del>
<del>cd JDE </del>
<del>mkdir -p workspace/mot16-2020-5-29 </del>
<del>./tools/split_dataset.sh ./workspace/mot16-2020-5-29 </del>
<del>此时workspace/mot16-2020-5-29目录下会生成train.txt </del>

<del>## 1.2 从预训练模型导出参数生成JDE初始模型<del>

<del>* 从darknet官网下载darknet53预训练模型 <del>
<del>wget https://pjreddie.com/media/files/darknet53.conv.74 -P ./workspace <del>
<del>python darknet2pytorch.py -pm ./workspace/mot16-2020-5-29/jde.pth \ <del>
<del> --dataset ./workspace/mot16-2020-5-29 -dm ./workspace/darknet53.conv.74 -lbo <del>
<del>此时workspace/mot16-2020-5-29目录下会生成初始模型jde.pth, 其骨干网已初始化为darnet53的参数 <del>

<del>## 1.3 执行训练脚本<del>

<del>cp ./tools/train.sh ./train.sh <del>
<del>根据需要修改, 然后运行训练脚本 <del>
<del>./train.sh <del>

<del># 2. 测试<del>
<del>运行类似如下命令执行多目标跟踪 <del>
<del>python tracker.py --img-path /data/tseng/dataset/jde/MOT16/test/MOT16-03/img1 \ <del>
<del> --model workspace/mot16-2020-5-29/checkpoint/jde-ckpt-049.pth<del>

JDE

JDE 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

JDE

JDE 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README