Practical overview
When it fits
Use it when you need a C++ or Python inference component for an exported YOLO model on NVIDIA hardware and can meet the CUDA and TensorRT prerequisites.
Who it is for
- Developers integrating YOLO inference into applications running on NVIDIA hardware.
- C++ and Python developers working with TensorRT-based computer-vision deployment.
Documented uses
- Run an exported YOLO model for object detection.
- Build inference applications for segmentation, classification, pose estimation or rotated-object detection.
- Use YOLO inference in a video-analysis application.
Requirements & limits
The public community edition is licensed under GPL-3.0. The repository describes it as version 6.4 and says its inference interface is frozen.
- The toolkit targets NVIDIA devices and depends on the CUDA and TensorRT environment.
- Model conversion and TensorRT engine building are assigned to the separate trtyolo-export package.
- Python bindings must be enabled during the project build.
- The community edition's inference interface is frozen.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Build and install TensorRT-YOLO
Compile the library and optionally its Python bindings from the repository.
Before you start
- NVIDIA environment with CUDA 11.0.1 or later and TensorRT 8.6.1 or later recommended.
- Linux x86_64 or ARM recommended; Windows is also supported.
- CMake; pybind11[global] is required when building Python bindings.
Steps
Clone the repository and change into its directory.
Install pybind11[global] if building Python bindings.
Run CMake configuration with TRT_PATH, CMAKE_INSTALL_PREFIX and, for Python bindings, BUILD_PYTHON=ON.
Build and install the configured targets with CMake.
Expected outcome
The install directory contains the library headers and dynamic libraries; enabling BUILD_PYTHON also generates Python binding files.
Guide limitations
- This setup guide covers compilation and installation, not model conversion or a completed inference run.
- The README recommends CUDA 11.0.1 or later and TensorRT 8.6.1 or later; verify requirements for the target environment.
Build a Python wheel from the Python-enabled build
Package the Python bindings as a wheel after compiling the project with Python support.
Before you start
- A TensorRT-YOLO build configured with BUILD_PYTHON=ON.
- The project's documented CUDA and TensorRT prerequisites.
Steps
Install or upgrade the Python build package with pip.
Run python -m build --wheel from the repository.
Install the generated dist/trtyolo-6.*-py3-none-any.whl file with pip.
Expected outcome
The generated wheel is installed into the Python environment.
Guide limitations
- The README presents this as packaging after enabling Python support; it does not by itself convert a model or demonstrate a complete inference run.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- laugh12321/TensorRT-YOLO · Official homepage title · retained in its source language