yolov5-trt

Minimal code for YOLO v5 inference with TensorRT (C++).

Tutorial: https://yinguobing.com/a-quick-tutorial-for-yolov5-tensorrt-cpp-inference/

Prerequisites

To build and run this program, the following items are required.

• a CUDA compatible GPU card
• NVIDIA driver: 470.82.01 or higher
• NVIDIA CUDA: 11.4.3
• NVIDIA cuDNN: 8.2.1
• NVIDIA TensorRT: 8.2.5.1
• OpenCV: 4.5.4
• cmake: 3.16 or higher
• an exported yolov5s TensorRT engine file

Install

Clone this repo:

git clone https://github.com/yinguobing/yolov5-trt.git

Build

Build with cmake:

cd yolov5-trt
mkdir build && cd build
cmake ..
make

Run

Run inference with a sample image:

yolov5 yolov5s.engine input.jpg

.. then checkout the output image.

Authors

Yin Guobing (尹国冰) - yinguobing

License

This project is licensed under the GPL-3.0 license - see the LICENSE file for details

Acknowledgments

• https://github.com/cyrusbehr/tensorrt-cpp-api