Yolov5 advanced seven target tracking latest environment construction (II)

I've already introduced deepsort The process of building the latest environment , This section is used github Another source code establishes the target tracking environment , This section provides information from Google Drive Download the crowd_human.pt As the weight file of recognition , At the same time, we can use class To limit the identified categories , Achieve some kind of tracking . The specific process is as follows ：
https://github.com/ynlx/Yolov5_DeepSort
This is a gihub On the other yolov5_DeepSort edition , Is in mike Fine tuned version on version , Easy to use ,tag Icon correspondence 3.0-5.0 edition , But note that the latest version still needs to climb over the wall . Establish the environment and requirements The process is not repeated , You can refer to Yolov5 Advanced seven target tracking the latest environment construction . Use Yolov5_DeepSort5.0 edition ,yolov5 Use 6.1.
If you are prompted No module named ‘yolov5.utils.downloads’ Namely yolov5 Outdated Version , Low version yolov5 utils Not in it downloads Folder .
Execute after installation

python track.py --source 0

After downloading pt The timeout failure will be prompted after the model .
After science goes online , On link pt After the document success , Test classification tracking

python track.py --source 0 --yolo_model yolov5s.pt --classes 16 17

The camera will capture yolov5 Medium 16 17 Category is cat and dog , Success after scientific Internet . Continue to test crowd tracking

python track.py --source 0 --yolo_model yolov5/crowdhuman_yolov5m.pt --classes 0

This crowdhuman_yolov5m.pt It's from google drive Downloaded , Should be yolo6.0 Later version . Download at https://download.csdn.net/download/buaaweibin/85853856

Therefore, we should avoid using scientific Internet ds3.0+ YOLO5.0 edition , The weight file needs yolo5.0 Environment self-study .