For simplified Chinese version: 简体中文版

This package provides a ROS wrapper for PyTorch-YOLOv5 based on PyTorch-YOLOv5. The package has been tested with Ubuntu 16.04 and Ubuntu 18.04.

V1.0.1: Add device options(cpu or gpu).

Authors: Zhitao Zheng ([email protected])

• Ubuntu 16.04 / 18.04
• ROS Kinetic / Melodic
• Python>=3.6.0 environment, including PyTorch>=1.7

bash ~/Downloads/Anaconda3-2021.05-Linux-x86_64.sh

export PATH=/home/your/anaconda3/bin:$PATH

source ~/.bashrc

conda create -n mypytorch python=3.8

conda activate mypytorch

Install PyTorch: https://pytorch.org/get-started/locally/

conda install pytorch torchvision cudatoolkit=10.2 -c pytorch

alias python='/home/your/anaconda3/envs/mypytorch/bin/python3.8'

source ~/.bashrc

cd /your/catkin_ws/src

git clone https://github.com/qq44642754a/Yolov5_ros.git

cd yolov5_ros/yolov5

sudo pip install -r requirements.txt

1. First, make sure to put your weights in the weights folder.
2. The default settings (using yolov5s.pt) in the launch/yolo_v5.launch file should work, all you should have to do is change the image topic you would like to subscribe to:

roslaunch yolov5_ros yolo_v5.launch

Alternatively you can modify the parameters in the launch file, recompile and launch it that way so that no arguments need to be passed at runtime.

• image_topic

  Subscribed camera topic.

• weights_path

  Path to weights file.

• pub_topic

  Published topic with the detected bounding boxes.

• confidence

  Confidence threshold for detected objects.