This repo is a ROS wrapper for the recently release Yolov7 architecture. Currently this project uses the Yolov7-mask architecture for segmentation of the detected images. This repo is based on the official implementation of the Yolov7 algorithm.

The ros node will take in images from a camera and output all the data including the masks, bounding boxes and the centers of the objects detected (in the camera frame) to a new topic. Visualization of the final result is also available on a separate topic but can be turned off if needed (WIP)

(back to top)

Following ROS packages are required:

• vision_msgs
• geometry_msgs
• sensor_msgs

Python requirements: The requirements.txt file needs to be installed and it can be done using the following command

pip install -r requirements.txt

Apart from this the Detectron 2 package is needed for processing the masks from the image.

(back to top)

Node parameters can be adjusted in either the launch file or through command line arguments while launching the launch file

roslaunch yolov7_ros yolo-v7.launch

(back to top)

• Add Ros Node
• Parameterize and add arguments
• Add paralellization to improve performance
• Add support for different model
• Morph into a more general Ros wrapper for Pytorch

(back to top)

Distributed under the MIT License. See LICENSE.txt for more information.

(back to top)

Aditya Rathi - [email protected]

Project Link: https://github.com/aditya-rathi/yolov7_ros

(back to top)

This project is based upon the Yolov7 official implementation and all credits for the algorithm lie with the authors of the [paper]((https://github.com/WongKinYiu/yolov7)

(back to top)