# tensorrt_inference

**Repository Path**: bewithmeallmylife/tensorrt_inference

## Basic Information

- **Project Name**: tensorrt_inference
- **Description**: this is project base on https://github.com/linghu8812/tensorrt_inference
- **Primary Language**: C++
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 1
- **Created**: 2022-11-01
- **Last Updated**: 2023-10-31

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# **TensorRT Models Deploy from ONNX Plus**


## This project is base on [TensorRT Models Deploy from ONNX](https://github.com/linghu8812/tensorrt_inference)

## What differences between this project and [TensorRT Models Deploy from ONNX](https://github.com/linghu8812/tensorrt_inference)
- Add mmpose HRNet whole body demo
- You can run this project in ready build docker image and don't need to  build docker image by yourself
- Add Town-Down Multiple Person Pose Estimation Interface With YOLOv7 and HRNet


## **How to Use**

- Clone the code
```shell
git clone https://gitee.com/bewithmeallmylife/tensorrt_inference.git
```

- Download ONNX models, and put `weights` folder into folder `/your-code-parent-path/tensorrt_inference`

[ONNX models](https://pan.baidu.com/s/1scqbfiDINinWMbTL8W94WA?pwd=46cz)

- Run docker container

```
sudo docker run --net=host  --gpus '"device=0"' --privileged  -it -v /your-code-parent-path/tensorrt_inference:/app/tensorrt_inference  -v /dev/video0:/dev/video0 -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY   bewithmeallmylife/tensorrt-inference-app:cuda10.2-cudnn8-trt7-devel-ubuntu18.04
``` 

- Build tensorrt_inference project in docker container:
```
cd /app/tensorrt_inference/project
mkdir build && cd build
cmake ..
make -j
```
after these steps, the executable program `tensorrt_inference` and all support model shared libraries have been built.

- Run tensorrt_inference

```
cd ../../bin
./tensorrt_inference
```
it will show:
```
Please design model arch, config file and folder name!
```

- Inference with models, take **yolov5** model as example, put `yolov5s.onnx` model into `weights` folder:

```
cd ../../bin/
./tensorrt_inference yolov5 ../configs/yolov5/config.yaml ../samples/detection_segmentation
```

Use one model library example can reference here [example](example)

## **Build from docker But you don't have to do this**

```
cd Dokerfile-runtime/cuda-10.2
sudo docker build -t='bewithmeallmylife/tensorrt-inference-runtime:cuda10.2-cudnn8-trt7-devel-ubuntu18.04' .
```
```
sudo docker build -t='bewithmeallmylife/tensorrt-inference-app:cuda10.2-cudnn8-trt7-devel-ubuntu18.04' .
```

## **Supported Models**



models|framework|instruction
---|---|---
[lenet](project/lenet)|PyTorch|An example from model training to TensorRT model deploy
[alexnet](project/alexnet)|MXNet Gluon|MXNet Gluon example
[arcface](project/arcface)|MXNet Symbol|MXNet Symbol and face recognition example
[CenterFace](project/CenterFace)|ONNX|rewrite ONNX model and face detection example
[efficientnet](project/efficientnet)|Keras|Keras to ONNX example
[face_alignment](project/face_alignment)|MXNet Symbol|MXNet Symbol and face key points  detection example
[fast-reid](project/fast-reid)|PyTorch|PyTorch and pedestrian reid example
[FCN](project/FCN)|GluonCV|MXNet GluonCV semantic segmentation example
[gender-age](project/gender-age)|MXNet Symbol|MXNet Symbol and face gender and age recognize example
[ghostnet](project/ghostnet)|PyTorch|PyTorch example
[MiniFASNet](project/MiniFASNet)|PyTorch|PyTorch face anti spoofing example
[mmpose](project/mmpose)|PyTorch|PyTorch person key points detect example
[nanodet](project/nanodet)|PyTorch|PyTorchlightweight anchor-free object detection example 
[RetinaFace](project/RetinaFace)|MXNet Symbol|MXNet Symbol and face detection example
[ScaledYOLOv4](project/ScaledYOLOv4)|PyTorch|YOLOv4 large with PyTorch implementation
[scrfd](project/scrfd)|PyTorch|PyTorch scrfd face detection example
[seresnext](project/seresnext)|PyTorch|PyTorch example
[Swin-Transformer](project/Swin-Transformer)|timm|timm image classification example
[yolor](project/yolor)|PyTorch|PyTorch and object detection example
[Yolov4](project/Yolov4)|darknet|darknet and object detection example
[yolov5](project/yolov5)|PyTorch|PyTorch and object detection example
[YOLOv6](project/YOLOv6)|PyTorch|PyTorch and object detection example
[yolov7](project/yolov7)|PyTorch|PyTorch and object detection example
 
 ## **Discussion Groups**

 - **QQ group:** 588966055