[start from scratch] detailed process of deploying yolov5 in win10 system (CPU, no GPU)

Preface

I also read others' blogs , Hit a lot of holes , It took nearly a morning to deploy . So please be clear before you start Corresponding version ！ Please follow the steps step by step . in addition to , Please download by yourself YOLOV5 Extract and save the source code ,GitHub - ultralytics/yolov5: YOLOv5 in PyTorch > ONNX > CoreML > TFLite

One 、Anaconda3 Download and install

Anaconda | The World's Most Popular Data Science Platform Enter official website , Don't download the latest version , Corresponding to the latest version at present Python3.9, The version is too new PyTorch Will also face the problem of updating , It's better to use a lower version ！！ Upper figure

 

 

  I downloaded it 2020 year 11 Of the month , Corresponding Python3.8.5 edition , Actually use 2021 year 05 No problem with the version . however Python The version should not be too new ！ The installation process is simple , I only have a few important pictures , All the unimportant pictures are dotted Next！！ No mistake .

Choose all here , The first represents automatic configuration of environment variables , The second represents the configuration Python3.8 edition . After installation, two options are cancelled （ Oh dear , What does it matter if you don't cancel ？ Not affecting use ）. thus ,Anaconda installation is complete ！

Two 、 utilize Anaconda3 download PyTorch

2.1 Find... In the lower left corner Anaconda Prompt

2.2 establish PyTorch Environmental Science

conda create --name pytorch pyrhon==3.8

 

  Next You also need to manually enter a “y"！！！！  Wait for installation

2.3 see pytorch Environment and activate

  obviously , except base Environmental Science , There is another one we created ourselves pytorch Environmental Science . Next , Activate pytorch Environmental Science .

conda activate pytorch

After activation , You can see the front base Turned into pytorch The environment .ok, Then let's install pytorch.

2.4 install PyTorch

Start Locally | PyTorch Enter the page and then come here , Select the configuration shown in the figure below （ ah ？ Why did you choose CPU ah ？ Because I don't have NVIDIA alone , So it doesn't work CUDA, If you have NVIDIA graphics card , Remember to choose CUDA！！） as for CUDA Part I ignored , Because I didn't install it , But I think it should be almost installed .

  Copy the red box in the above figure , namely

conda install pytorch torchvision torchaudio cpuonly -c pytorch

  In what we have just entered PyTorch Enter in the environment

  This process also You need to manually enter a “y"！！！ If during your installation Not all downloads succeeded , then Repeat several times This Directive ！！

2.5 verification PyTorch Is the installation successful

Activate first pytorch Environmental Science , Reuse python, Second tone torch library , The final output torch Version number of ！

import torch
torch.__version__

  If you can do this smoothly , You can read Chapter 3 directly . however ！！！ Not everyone is plain sailing , For example, I , I can enter here Import torch Report errors ：No module named "torch"！！ I tried many ways , Finally, I changed my mind ： Check on pytorch Is it really installed in the environment pytorch.

Enter into pytorch Folder for environment , Enter... In the menu bar cmd

 

  Type again

conda install pytorch torchvision torchaudio cpuonly -c pytorch

  then , I'll test it again , It's wrong not to report ！

3、 ... and 、 stay PyTorch Install in Opencv

This is an easy step , Enter the first Anaconda Prompt Console , Activate PyTorch Environmental Science , download opencv that will do .

pip install python-opencv

Remember to test after completion opencv

import cv2
cv2.__version__

without . Then it's the same as the end of Chapter 2 , First of all to enter pytorch Environmental Science , Execute it again cv Installation instructions are sufficient .

pip install python-opencv

Four 、PyCharm Environment building

  Skip here PyCharm Download and install .

4.1 Open it up and download okay detect.py file

 4.2 Set compilation configuration

choice detect.py

4.3 Set interpreter

  Here we set the settings in Chapter 2 pytorch frame .

5、 ... and 、 The test results

When everything is ready , Click on the compilation , Miracles happen ！ If you appear No module named "XXXXX", Please read on .

And in runs The test results can be seen in the folder ... 

 

6、 ... and 、 Rely on the supplementary installation of the Library

First , stay Anaconda Prompt Active in pytorch Environmental Science

conda activate pytorch

secondly , Install the dependent libraries in turn , What's missing for installation . for example , The lack of pandas library , be ：

conda install pandas