ros_yolov7's Introduction

ros yolov7 bounding boxes + body keypoints

Ros package publishing ros yolov7 output tested with ros noetic but should would with others ros versions aswell

Based on https://github.com/RizwanMunawar/yolov7-pose-estimation

Follow the tutorial to install yolov7 required packages

Configure launch file:

/topic_img for your input topic
/path_yolo_model path to yolo weight downloaded following https://github.com/RizwanMunawar/yolov7-pose-estimation tutorial
use launch-prefix if you want to start the node with a python env, else remove it

start with:

roslaunch ros_yolov7 pose_estimate.launch

An example of subscriber is in src/ directory

Packages installations

cd /src
pip install -r requirements.txt
pip install --extra-index-url https://rospypi.github.io/simple/ rospy rosbag rospkg
pip install scikit-learn==0.22.2 --upgrade

ros module documentation

launching the module with Docker

to launch the module independently, the following docker-compose command must be run (**from the parent directory of the catkin workspace) : docker-compose -f catkin_ws/src/ros_yolov7/docker-compose.yml --env-file catkin_ws/src/ros_yolov7/.env -d it is also possible to run the command docker-compose up -d from this directory

formats used

BboxKp

int64 id : a user's id
float64[] bbox : a user's bounding boxes
float64[] kp : a user's keypoints

BboxKpList

BboxKp[] subjects : a list of BboxKp

ROS topics

subscribed to

[topic_img] (the parameter defined in the launch file) : Image, the image stream feeded to the service

published

/refined_perception/kp_bbox : BboxKpList, the users' bounding boxes and keypoints, useful for identifying them and interpret their position/body posture

Recommend Projects