yolov5 environment configuration series tutorials under Windows11 (problem analysis part)

An error occurred when installing cudnn

insert image description here
It can be seen that we reported an error when we installed cudnn, and this error can be solved by changing the source.

Solutions to errors when installing cudnn
- 1. Find the cudnn version
and enter the following code

anaconda search -t conda xxxx（你要下载的软件名）

insert image description here

Find the version we need in a lot of versions it gives
(the above cudnn and cuda correspondence table can see the required version)
insert image description here
record the version name

insert image description here
-2. Download the version you need
and enter the following code

conda install -c https://conda.anaconda.org/X (your version name)

conda install -c https://api.anaconda.org/fastchan cudnn
								 		  （版本名中的/用空格代替）

just download

-3. Check if the download is successful
and enter the following code

conda list

insert image description here
You can see that there is one more cudnn in our environment

4. Verify that CUDA and duDNN can be used normally
as follows

python
import torch
a = torch.tensor(1.)
a.cuda()
from torch.backends import cudnn
cudnn.is_available()
cudnn.is_acceptable(a.cuda())

insert image description here

Two problems encountered in configuring the Yolov5 environment

insert image description here
The installation with the official path always fails, so you need to replace the domestic path.
The ideal is the mirror image of Douban, as follows

pip install -r requirements.txt -i https://pypi.douban.com/simple --trusted-host pypi.douban.com

The specific implementation process is as follows

insert image description here
But at the end of the execution, there is still an error

insert image description here
The key to the error statement is :
error : Microsoft Visual C++ 14.0 or greater is required. Get it with “Microsoft C++ Build Tools”: https://visualstudio.microsoft.com/visual…

Solution : Install the Microsoft Visual C++
offline package. Here I will give you a link to the Baidu network disk. You can download it
yourself . Slow, please be patient! After the installation is complete, go back to cmd and run the command:

insert image description here

pip install nuscenes-devkit -i http://pypi.douban.com/simple --trusted-host pypi.douban.com

insert image description here
You can succeed! ! !

Three problems encountered in running the test

The ultimate solution to the "No module named 'cv2'" error after running with Pycharm. I
just configured the environment of python and opencv today, and found that OpenCV can be successfully used on PowerShell, and no error appears after entering import cv2, as shown below:

insert image description here

Python reports No module named 'cv2'
import cv2;
Python reports No module named 'cv2' during operation

Reason : The opencv-python library is not installed
Command line mode installation: pip3 install opencv-python

JamSonsMacBook-Pro:~ jamson$ pip3 install opencv-python
Collecting opencv-python
  Downloading opencv_python-4.2.0.32-cp37-cp37m-macosx_10_9_x86_64.whl (47.9 MB)
     |▊                               | 1.0 MB 22 kB/s eta 0:34:31

If in PyCharm, the installation method is as follows:

insert image description here
Search opencv-python, click [Install Package]
when the installation is complete, it will prompt "Package 'opencv-python' installed successfully"

Open PyCharm, select file-settings-project-project interpreter, as shown below. If you don't see the opencv-python package after opening it, select the "+" sign in the upper right corner to search for opencv-python. Finally, remember to click the application in the lower right corner.

insert image description here
The above are some of the problems I have encountered, I hope to help you. Welcome everyone to discuss together, learn from each other and make progress together!