Running YoloV5 with TensorRT Engine on Jetson Orin Nano 8GB.

This repository contains step by step guide to build and convert YoloV5s model into a TensorRT engine on Jetson Orin Nano 8GB.

Please install Jetpack OS version 5.1.1 as mentioned by Nvidia and follow below steps. Please follow each steps exactly mentioned in the video links below :

Build YoloV5 TensorRT Engine on Jetson Nano: https://www.youtube.com/watch?v=ErWC3nBuV6k

Object Detection YoloV5 TensorRT Engine on Jetson Nano: https://www.youtube.com/watch?v=-Vu65N1NRWw

Please take note that this repo only work for yolov5s, i got some error when try to convert yolov5m from wts to engine, but you can refer to https://github.com/wang-xinyu/tensorrtx/tree/yolov5-v7.0/yolov5 for futher exploration.

Jetson Xavier:

Install Libraries

Please install below libraries::

sudo apt-get update
sudo apt-get install -y liblapack-dev libblas-dev libatlas-base-dev gfortran libfreetype6-dev libopenblas-base libopenmpi-dev libjpeg-dev zlib1g-dev
sudo apt-get install -y python3-pip

Install below python packages

Numpy comes pre installed with Jetpack, so make sure you uninstall it first and then confirm if it's uninstalled or not. Then install below packages:

numpy==1.19.0
pandas==0.22.0
Pillow==8.4.0
PyYAML==3.12
scipy==1.5.4
psutil
tqdm==4.64.1
imutils

You can put all these libraries into a req.txt. Then run, pip3 install -r req.txt

Install PyCuda

We need to first export few paths. To install tensorrt into python3, you can refer to https://docs.nvidia.com/deeplearning/tensorrt/install-guide/index.html#maclearn-net-repo-install.

export PATH=/usr/local/cuda/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
sudo apt-get install python3-libnvinfer
python3 -m pip install pycuda --user

Install Seaborn

$ sudo apt install python3-seaborn

Install torch & torchvision

wget https://github.com/KumaTea/pytorch-aarch64/releases/download/v1.11.0/torch-1.11.0-cp38-cp38-linux_aarch64.whl
pip3 install torch-1.11.0-cp38-cp38-linux_aarch64.whl
git clone --branch v0.13.0 https://github.com/pytorch/vision torchvision
sudo chmod -R 777 /usr/local/lib/python3.8/dist-packages/ 
cd torchvision
python3 setup.py install 

Not required but good library

python3 -m pip install -U jetson-stats==3.1.4

This marks the installation of all the required libraries.

---

Generate wts file from pt file

Yolov5s.pt and Yolov5n.pt are already provided in the repo. But if you want you can download any other version of the yolov5 model. Then run below command to convert .pt file into .wts file

cd JetsonYoloV5
wget https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5s.pt
python3 gen_wts.py -w yolov5s.pt -o yolov5s.wts

Make

Create a build directory inside yolov5. Copy and paste generated wts file into build directory and run below commands. If using custom model, make sure to update kNumClas in yolov5/src/config.h

cd yolov5/
mkdir build
cd build
cp ../../yolov5s.wts .
cmake ..
make 

Build Engine file

cd yolov5/src
cat config.h
./yolov5_det -s yolov5s.wts yolov5s.engine s

Testing Engine file

./yolov5_det -d yolov5s.engine ../images

This will do inferencing over images and output will be saved in build directory.

---

Python Object Detection

Use app.py to do inferencing on any video file or camera.

$ python3 app.py

If you have custom model, make sure to update categories as per your classes in yolovDet.py .