PolySTAR - CV Robomaster 2026

Work repository for the Computer Vision team

Introduction

This repository contains the work of PolySTAR's computer vision team for the 2026 RoboMaster competition, organized by DJI. Based on last year's repo.

For a step by step example of how to run the code on Jetson, please check here

Repo organization

• doc/ : General & specific documentation about our work. Everything should be documented ;)
• ros_ws/ : ROS workspace. This directory contains everything needed to run the detection
• detection/ : CV dataset pre-processing & training

Requirements

ROS

• ROS 2 Iron
• OpenCV & Python bindings

Detection

This project relies on the darknet library to train/test the model as well as the deepstream library to run on Jetson. darknet is included as a submodule and deepstream is already installed on the Jetson. They are both required to build the ROS workspace.

To setup darknet, please check here. To train/test the model with darknet on computer, please check here.

If deepstream is already installed, you still need to add up a wrapper to run darknet model, check here

Code guidelines

Please format your code before commiting :

• For C++ code, a .clang-format file is at the root of the ROS workspace to ensure consistency.
• Python code is formatted using autopep8

Datasets & Training

TODO