YOLO5 construction

This tutorial is suitable for building the official image of Jetson Orin nx yourself. If you are using the YAHBOOM version of the image, this tutorial can be ignored.

1.preparation

jetson orin nx *1 jetson orin nx configuration is as follows：

If you have read the installation tutorials for torch and torch vision in the previous chapter, you can start directly from section 2.3

2.Environment construction of YOLO5（yolo5 v5.0）

2.1 Download the required module for YOLO5- torch

（If you follow the tutorial to install the Jetson-reference environment, this section can be ignored）

sudo apt-get install python3-pip libopenblas-base libopenmpi-dev 
pip3 install Cython
pip3 install numpy torch-1.12.0a0+2c916ef.nv22.3-cp38-cp38-linux_aarch64   
sudo apt-get install libjpeg-dev zlib1g-dev libpython3-dev libavcodec-dev libavformat-dev libswscale-dev

torch-1.12.0a0+2c916ef.nv22.3-cp38-cp38-linux_aarch64 This file is obtained from the attachment built in the environment and transferred to Jetson through WinSCP

2.2 Install the corresponding version of torchvision

xxxxxxxxxx
git clone --branch v0.13.0 https://github.com/pytorch/vision torchvision
cd torchvision
export BUILD_VERSION=0.13.0
python3 setup.py install --user

If git clone reports an error, please check the network and run again

2.3 Download the source code of YOLO5

xxxxxxxxxx
cd ~
git clone https://github.com/marcoslucianops/DeepStream-Yolo
python3 -m pip install --upgrade pip 
cd yolov5

Because Jetson orin nx already comes with opencv4.5.4, there is no need to install Python's opencv. You can verify it through import cv2 Therefore, we need to open the requirements file in the yolo5 directory and add a # sign before this line Run after modification

xxxxxxxxxx
pip3 install -r requirements.txt -i https://mirror.baidu.com/pypi/sample

Wait until the download is complete

3.Verify if yolo5 has been successfully built

xxxxxxxxxx
cd ~/yolov5
python3 detect.py

Wait for him to automatically download the weight file. If the network is not working, please obtain the yolov5s.pt file from the attachment provided by us and place it under the yolov5 folder If there are no errors reported, it indicates that YOLO5 has been successfully built and the identified results will be stored in the YOLOv5/runs/detect/exp path

The following is a picture of successful operation Exp5: Because the command python3 detect.py was run for the fifth time, the results were stored in the directory of exp5

If you run the third step, there may be an error due to the high version of the torch, so we need to make some modifications

xxxxxxxxxx
sudo gedit /usr/local/lib/python3.8/dist-packages/torch/nn/modules/upsampling.py 

Make a modification to line 153, recompute_scale_factor=self.recompute_scale_factor Just add '#' in front of the factor line. The modified results are shown in the following figure： Run step 3 again and it will work properly.