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
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
- int64 id : a user's id
- float64[] bbox : a user's bounding boxes
- float64[] kp : a user's keypoints
- BboxKp[] subjects : a list of BboxKp
- [topic_img] (the parameter defined in the launch file) : Image, the image stream feeded to the service
- /refined_perception/kp_bbox : BboxKpList, the users' bounding boxes and keypoints, useful for identifying them and interpret their position/body posture