一、项目介绍：

二、环境安装

• 平台：windows 10
• 编译器：pycharm
• cuda 11.3
• cudnn 8.2.0.53

conda create -n yolov5bytetrack  python=3.7
conda activate yolov5bytetrack 
pip install  prettytable -i https://mirror.baidu.com/pypi/simple
pip install ujson -i https://mirror.baidu.com/pypi/simple
pip install opencv-python==4.6.0.66 -i https://mirror.baidu.com/pypi/simple
pip install pillow -i https://mirror.baidu.com/pypi/simple
pip install tqdm -i https://mirror.baidu.com/pypi/simple
pip install PyYAML==5.1 -i https://mirror.baidu.com/pypi/simple
pip install visualdl==2.2.0 -i https://mirror.baidu.com/pypi/simple
pip install scipy==1.0.0 -i https://mirror.baidu.com/pypi/simple
pip install scikit-learn==0.21.0 -i https://mirror.baidu.com/pypi/simple
pip install gast==0.3.3 -i https://mirror.baidu.com/pypi/simple
pip install faiss-cpu -i https://mirror.baidu.com/pypi/simple
pip install easydict -i https://mirror.baidu.com/pypi/simple
pip install paddlepaddle-gpu==2.4.2.post117 -f https://www.paddlepaddle.org.cn/whl/windows/mkl/avx/stable.html
pip install psutil
pip install seaborn
pip install paddleclas -i https://mirror.baidu.com/pypi/simple
pip install loguru
pip install thop
conda install -c conda-forge lap
pip install mediapipe

三、执行代码

1、训练模型

下载数据集，文章百度云盘有提供

用的是PaddleCls进行训练，下载PaddleClas

git clone https://github.com/PaddlePaddle/PaddleClas

将下载的数据集解压，放到PaddleClas的dataset目录。

找到ppcls/configs/PULC/person_attribute/PPLCNet_x1_0.yaml配置文件，配置图片和label路径。

DataLoader:
  Train:
    dataset:
      name: MultiLabelDataset
      image_root: "dataset/pa100k/"                     #指定训练图片所在根路径
      cls_label_path: "dataset/pa100k/train_list.txt"   #指定训练列表文件位置
      label_ratio: True
      transform_ops:

  Eval:
    dataset:
      name: MultiLabelDataset
      image_root: "dataset/pa100k/"                     #指定评估图片所在根路径
      cls_label_path: "dataset/pa100k/val_list.txt"     #指定评估列表文件位置
      label_ratio: True
      transform_ops:

检测数据集格式是否正确，train_list.txt的格式为

00001.jpg    0,0,1,0,....

配置好后，就可以直接训练了

python tools/train.py -c ./ppcls/configs/PULC/person_attribute/PPLCNet_x1_0.yaml

训练完后，导出模型

python3 tools/export_model.py -c ./ppcls/configs/PULC/person_attribute/PPLCNet_x1_0.yaml -o Global.pretrained_model=output/PPLCNet_x1_0/best_model -o Global.save_inference_dir=deploy/models/PPLCNet_x1_0_person_attribute_infer

将导出的结果放在deploy/models/PPLCNet_x1_0_person_attribute_infer/目录下

便可以使用PaddleCls提供的函数直接调用模型，测试模型效果

import paddleclas

model = paddleclas.PaddleClas(model_name="person_attribute")
result = model.predict(input_data="./test_imgs/000001.jpg")
print(result)

python tools/infer.py 
    -c ./ppcls/configs/PULC/person_attribute/PPLCNet_x1_0.yaml
    -o Infer.infer_imgs=../Butterfly20/001.Atrophaneura_horishanus/084.jpg 
     -o Global.pretrained_model=./output/PPLCNet_x1_0/latest

输出结果如下：

[{'attributes': ['Female', 'Age18-60', 'Front', 'Glasses: False', 'Hat: False', 'HoldObjectsInFront: True', 'ShoulderBag', 'Upper: ShortSleeve', 'Lower:  Trousers', 'No boots'], 'output': [0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 0, 0], 'filename': './test_imgs/000001.jpg'}]

2、运行yolov5

conda activate yolov5bytetrack 
pip install lap

3、ByteTrack

cython_bbox库就该这么安装

pip install -e git+https://github.com/samson-wang/cython_bbox.git#egg=cython-bbox

四、效果展示

五、总结