#5757th of 71 in Model servingFor science, health and research, robotics
yolo_ros
ROS 2 node running Ultralytics YOLO ONNX models for detection, pose and segmentation
- Stars
- 1.2k
- License
- MIT
- Last commit
- Nov 2026
- Last release
- Sep 2026
- Language
- C++
993 on 21 Apr 20261,159 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
C++ ROS 2 package that runs Ultralytics-exported YOLO ONNX models through ONNX Runtime, with no Python or ultralytics dependency at run time. A single yolo_node loads detection, tracking, 3D lifting from depth images and debug rendering as pluginlib plugins, configured from one YAML file. It supports detect, segment, pose, OBB and classify tasks on CPU, CUDA or TensorRT.
Who it is for: Robotics engineers adding YOLO perception to ROS 2 robots
Strengths
- Pure C++ and ONNX Runtime; no Python runtime or ultralytics install needed at run time
- One process with ordered plugin chain (detection, tracking, 3D, debug) configured by one YAML
- Covers detect, segment, pose, OBB and classify; YOLOv3 through YOLO26 models
- CPU on x64 and aarch64, CUDA, TensorRT, plus a QNN Hexagon NPU path
Weaknesses
- Ultralytics weights are AGPL-3.0, so the MIT license does not cover the models
- Prebuilt GPU ONNX Runtime is x64-only and limited to CUDA 11, 12 and 13
- QNN NPU support requires building ONNX Runtime from source
- Needs a ROS 2 colcon workspace; not usable outside ROS 2
What it needs
- GPU optional
- Docker
- Needs ROS 2, colcon, ONNX Runtime 1.20.0 (vendored), CUDA and cuDNN (GPU only), TensorRT 10 (optional)
- Models: YOLOv3, YOLOv5, YOLOv8, YOLOv9, YOLOv10
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Model serving
See all 71| Rank | Project | Score |
|---|---|---|
| 1 | llama.cppC/C++ inference engine for GGUF LLMs and VLMs with an OpenAI-compatible server | 94 out of 100 |
| 2 | OllamaLocal model runner with CLI and REST API for open-weight LLMs | 92 out of 100 |
| 3 | LocalAISelf-hosted inference server with OpenAI-compatible API and pluggable backends | 91 out of 100 |
| #4 | vLLMInference engine serving LLMs with PagedAttention and an OpenAI-compatible API | 88 out of 100 |
| #5 | strataRuns a 125B-parameter MoE model on one consumer GPU plus system RAM | 88 out of 100 |