search

atom-robotics-lab / ros-perception-pipeline

This is a ROS 2 based plug-and-play perception pipeline which can be customised to your need
https://atom-robotics-lab.github.io/
8 stars 9 forks source link

Re-wrote YOLOv5 Plugin using torch instead of OpenCV #59

Closed topguns837 closed 9 months ago

topguns837 commented 9 months ago

Description 📖

The code for YOLOv5 Plugin was originally written using OpenCV. This made it difficult to run YOLOv5 inference using GPU, since OpenCV had to be built locally. This code has been converted to torch which makes the codebase easy to understand and future-proof

Fixes # https://github.com/atom-robotics-lab/ros-perception-pipeline/issues/57

Type of change 📜

Please delete options that are not relevant.

How Has This Been Tested? 👀

Tested inside the given docker setup. I have used the KITTI Dataset as a source of video feed

Test Configuration 🖥️