Deepsort yolov5 in ros melodic1

Environment

• Ubuntu 18.04, cuda 10.2, RTX 2070 super, Ros melodic 1

Associative Repository : How to use ROS

Direction

If you created an ROS workspace(ex. catkin_ws), you put the git clone https://github.com/winston1214/Deepsort_yolov5_ros.git in workspace

$ cd catkin_ws/src
$ git clone https://github.com/winston1214/Deepsort_yolov5_ros.git
$ cd ..
$ catkin_make

Train

YOLOv5 Training

• Simple weight file(small_best.pt)

  : pretraining weight model - yolov5s.pt(As the size of the weight model increases, real-time performance decreases.)

  : batch size - 8, epochs - 150 ,classes - person, car

Detect

• ROS Image Topic Subscriber

$ roscore
$ cd catkin_ws/src/Deepsort_yolov5_ros
$ python3 ros_track.py --source ros --weights weights/small_best.pt

• Image, Video, webcam etc..

$ python3 ros_track.py --source file_path --weights weights/small_best.pt