Yolov5 advanced zero environment rapid creation and testing

After the traditional image processing course , Neural network is the inevitable choice , Was still in tensorflow Framework applicable yolov2, current v5 It is said that there has been a great leap forward , Each verified click on the website actually provides a large amount of accurate image recognition data to the corresponding company .
build yolo It mainly follows the following steps
1 install anaconda A little , I also installed pycharm Prefer in pycharm Programming .
2 Create an environment , First, create a folder in a hard disk space dedicated to various environments , Used to save the environment , Save scripts and data . I am here D disc pycharm Created under the project folder , How to apply it will be discussed later pycharm.

conda create --prefix=D:/PycharmProjects/yoloV5/yoloV5_env python=3.9

The above command is in yolov5 Under the document , Create a yoloV5_env Save environment , hinder python The version should say , use conda Management is the benefit , Don't go to base Call interpreter , No matter what environment is installed in the future , All of them should correspond to python Put the version interpreter into the environment .

conda activate D:\PycharmProjects\yoloV5\yoloV5_env

Activate this environment , And then install it online , Very fast

pip install yolov5

There should be no errors or warnings during installation , And then in python In the environment import yolov5 and import torch success
Retest cuda, You can't use the graphics card to speed up without installation

torch.cuda.is_available()
False

My computer is a notebook , It's not very useful , I won't discuss it here .

Let's put the environment in pycharm management
open pycharm

Choose carefully conda Environmental Science , interpreter The interpreter will search for unmanaged yoloV5 Environmental Science py3.9 Interpreter
Then it will prompt that this is an existing environment , And then you can do it in pycharm To manage the environment and write programs

Of course, the weight file is missing ,pt Download the files and test images by yourself , It can also be in github Download the yolo Look in the bag .

import yolov5
model = yolov5.load('yolov5s.pt')
img = 'images/zidane.jpg'
results = model(img, augment=True)

predictions = results.pred[0]
boxes = predictions[:, :4] # x1, x2, y1, y2
scores = predictions[:, 4]
categories = predictions[:, 5]

results.show()

The above program can realize the image recognition test

Although only installed CPU edition , But it is also suitable for ordinary notebooks , And the process is not very simple .