tensorrtOnnx

Implementation of some methods or tasks with TensorRT and ONNX.

Requirements

• TensorRT >= 7.0.0.11
• OpenCV >= 3.4.0
• yaml-cpp
• cuda

Changelog

Please refer to changelog.md for details and change history.

Getting Start

First clone this repo to your workspace, and enter this folder, then execute commands following below. After the executable file is generated, run it.

Note

You need to modified the cfgs/main.yaml to turn on the task you want to excute and prepare corresponding model following tutorials.

$ cp ${ONNX_MODEL_PATH} models
$ mkdir build && cd build
$ cmake .. && make -j
$ ./engine

Performance

# Waiting for update.

Notes

• All parameters can change in cfgs/*.yaml and cfgs/tasks/*.yaml. You can change it as your requirement.
• If you deploy this project on NX or Nano devices, you may need to modified the compute arch in CMakeLists.txt, more details in Nvidia official webs.
• If you deploy this project on GPU server, make sure you install CUDA and Cudnn with corresponding version with TensorRT, and modified {TRT_ROOT} add cuda include directory in CMakeLists.txt.