Target detection - make your own deep learning target detection data set with labelimg

Hello everyone , I meet you again , I'm your friend, Quan Jun .

Catalog

1 labelimg Introduce

2 labelimg Installation

3 Use labelimg

3.1 Data preparation

3.2 Some settings before annotation

3.3 Start tagging

---

1 labelimg Introduce

Labelimg It's an open source data Dimensioning tools , Three formats can be marked .

1 VOC Label format , Save as xml file .

2 yolo Label format , Save as txt file .

3 createML Label format , Save as json Format .

2 labelimg Installation

This is mainly about window Installation in the system , Start by opening cmd Command line （ Shortcut key ：win+R）. Get into cmd Command line console . Enter the following command ：

pip install labelimg -i https://pypi.tuna.tsinghua.edu.cn/simple

Copy

After running the above command , The system will automatically download labelimg Dependent dependency . Because this is a very lightweight tool , So it downloads quickly , When the in the following red box tells us that the installation is successful , explain labelimg Installation is successful .

3 Use labelimg

3.1 Data preparation

First of all, we need to prepare the data set that we need to mark . Here I suggest to create a new one named VOC2007 Folder （ This is a convention , It's OK not to do so ）, Create a new one called JPEGImages The folder of is used to store the image files that we need to label ; Create another one called Annotations Store labeled label files ; Finally, create a file named predefined_classes.txt Of txt File to store the category name to be marked .

VOC2007 The directory structure of is ：

├── VOC2007 │├── JPEGImages Store image files that need to be labeled │├── Annotations Store labeled label files │├── predefined_classes.txt Define all categories you want to label （ This document is optional , But when we define more categories , It's best to have this to create such a txt File to store categories ）

3.2 Some settings before annotation

First, in the JPEGImages This folder places the pictures to be marked , Here are three kinds of pictures , They are people 、 Dogs and cats .

And then again predefined_classes.txt This txt Enter the defined category in the document ; As shown in the figure below .

open cmd Command terminal （ Shortcut key ：win+R）. Enter the newly created VOC2007 route （ This is very important , It involves whether we can use predefined_classes.txt This txt Categories defined in the document , I've been stuck here for a long time , Once thought it couldn't show txt The category defined in the file is that I have problems installing ）. Execute the command shown in the figure to enter VOC2007 Under the path （ Everyone's path is different , Write according to your personal path ） As shown in the figure below ： You can see that you have entered the corresponding directory .

Enter the following command to open labelimg. This command means to open labelimg Tools ; open JPEGImage Folder , initialization predefined_classes.txt The classes defined inside .

labelimg JPEGImages predefined_classes.txt

Copy

Running the above command will open the tool ; as follows .

The buttons we commonly use in the figure are introduced below .

The path folder of the picture data to be marked , When you enter the command here, you select JPEGImages.（ Of course, this can be changed ）

The path folder where the category label is saved , Here we have chosen Annotations Folder .

This button can indicate that the label we marked is voc Format , Click to change to yolo perhaps createML Format .

Click on View, The options in the red box as shown in the figure... Will appear . You'd better tick it like me .

Auto Save mode： When switching to the next picture , The label is automatically saved .

Display Labels： The callout box and label... Are displayed

Advanced Mode： The marked cross will remain suspended in the window .

Common shortcut keys are as follows ：

A： Switch to the previous picture

D： Switch to the next picture

W： Call up the marked cross

del ： Delete dimension frame

Ctrl+u： Select the annotated picture folder

Ctrl+r： Select the marked label The folder where the tag exists

3.3 Start tagging

Because the cross we set for annotation is always on the annotation interface , This does not require us to press the shortcut key w, Then select the object we need to label . Hold down the left mouse button and drag it out of the box . As shown in the figure below , When we choose our goal , It will be loaded predefined_classes.txt Define all categories you want to label （ If there are many categories , It's really convenient , You don't need to type the name of each category by yourself ）. The category of the frame will be on the labeled frame （ The picture is not very clear due to color , Look carefully and you will find ）. Then the typed category label will appear on the far right of the interface . After typing a picture , Shortcut key D, Will enter the next , At this time, the label file will be automatically saved （voc Format will be saved xml,yolo Will save txt Format ）.

After the label is typed, you can go to Annotations Under the file, you can see that the tag file has been saved in this directory .

Since then labelimg The use explanation is over .

Publisher ： Full stack programmer stack length , Reprint please indicate the source ：https://javaforall.cn/147722.html Link to the original text ：https://javaforall.cn