YOLOv5_engine (TensorRT, YOLOV5 6.1, .engine)

This project is used to deploy the engine model (.engine) generated by YOLOV5 6.1 (export.py) to jetsonnano or running in a separate environment. You only need to install the tensorRT framework without torch

You only need to install tensorrt and pycuda for model(.engine) forward prediction.

Preparation

• My environment

CUDA: 11.3.1
cudnn: 8.2.1
python:3.9

• Installation If you use a Linux based system, you can easily install by:

  pip install -r requirements.py

else if you use a Windows system, only the installation method of tensorrt library is different, please reference by follow url https://docs.nvidia.com/deeplearning/tensorrt/install-guide/index.html#installing-tar

Usage

• Prediction support 2 MODE（--source camera or image) You can easily start by following command.

  python yolov5pred.py 

Other optional extra parameters：

--engine runs/yolov5s.engine  # Trained engine file path
--categories runs/names.txt   # Network prediction class file
--source camera               # camera or image for camera prediction or single image prediction
--conf-thres 0.25             # confidence threshold
--iou-thres 0.1               # iou threshold

Result

Prediction results of Windows or Linux platforms: