WesleyCh3n / trt-opencv

An example showing pipeline between OpenCV and TensorRT inference, and some pre/post-process utilities.

[T]ensor[RT] with OpenCV

libtrt: A example wrapper for inferencing using TensorRT with OpenCV GpuMat as input.

libdnn: Class Yolov8 and Resnet with dynamic batch size model based on libtrt to predict OpenCV GpuMat(s).

Tested Environment

• Ubuntu 22.04
• CUDA 12.3
• Cudnn 8.9.4.25
• TensorRT 8.6.1.6
• OpenCV 4.8.0 with CUDA enable

Quick start

Building Library

cmake -B build .
cmake --build build --parallel

Install Library

sudo cmake --install build --prefix /opt/trt # use prefix to install where you want

Example

build/yolov8_single model/yolov8.trt 20231212-161000.740_00_126.jpg output.jpg
build/resnet_single model/512-model-fp16.trt 20240114-041024.091_00_126_6_0.jpg output.jpg