When SSD (a recognition algorithm, not the solid state hard disk hhh) is involved in image detection, the way of labeling has changed. For the previously used caffe framework, you can write a label directly after the image name in the txt, but you need to create an xml file in the SSD. What is stored in this xml? In fact, you can open the file in the form of text, which stores the coordinates of the specific area in the image and the label you put on the picture, which is also a label file.
Ok, let’s not talk too much nonsense, let’s enter the topic, how to download labelImg under windows10~*(ai! How long does this lowercase L and uppercase i look the same...)
1. Download of labelImg
The method introduced in this article uses Anaconda3 , but there are many online tutorials for downloading Anaconda3 , so I won't repeat it in this article. Friends who have not downloaded it can download it and then come back to this tutorial.
1. Download some installation packages in Anaconda Prompt
#注意：大小写一定要区分，完全按照下面的输入
pip install PyQt5
pip install pyqt5-tools
pip install lxml #在正常下载anaconda3的情况下,lxml应该已经具备
pip install labelImg Some people may not know where Anaconda Prompt is? It is actually a python terminal. Open the "Start" in the lower left corner, the location is as shown in the figure ( the third one in the Anaconda3 (64-bit) directory ):
2. Reopen Anaconda Prompt and enter IabelImg
The previous download may take a while, but just install the above installation packages in order. Enter labelImg again, like the following: The
following box will pop up:
So far, labelImge is downloaded! ! ! Let's talk about how to use this software!
Second, the use of labelImg
1.labelImg interface introduction
In fact, labelImg is to label the designated area of the designated picture and store it in the xml file. First understand its working interface!
Here I introduce some working keys, because I use a few (hey, I don’t know what the others are for~)
The following format is the figure above and the description below.
The function of open is to open a picture, and the opened picture will be displayed in the middle work area, like the following:
open dir is to facilitate batch processing, you can select a picture folder, after selection, the picture list in the folder will be in The File List in the lower right corner shows, like this: The
generated xml file always needs a location to save, and it is through this setting the save path.
When importing in list form, click this to replace the next one.
When importing in list form, click this to return to the previous one.
Save to save the currently generated xml file
needs to frame the specified area, Create\nRectBox does this, Delete\nRectBox is to delete the frame (the second one, I have never used it, and I don’t know~)
2. Method of labeling
Marking is also relatively easy, according to the method of the previous frame, select the designated area. The following box will pop up:
Enter the name of the label you want to make in the box.
3. Commonly used shortcut keys
Do I want to point one by one for hundreds of pictures? Is that hand gone? Of course, labelImg has shortcut keys that can speed up our work efficiency
1.ctrl+s 作用：保存生成的xml文件
2.w 作用：框矩形
3.d 作用：换下一张
4.a 作用：回到上一张4. Note:
In the whole process of using labelImg, Anaconda prompt cannot be closed, there will be file records, As shown below:
In fact, the use of this software is relatively easy to learn, and Duoduo can quickly master it by practicing it by himself.
How could I directly meet such a complicated thing? Naturally, references to excellent information on the Internet are indispensable. Thanks to the uploader:
Installation of labelImg: https://jingyan.baidu.com/article/5225f26ba428fee6fa090829.html link
Use of labelImg: https://blog.csdn.net/qq_34108714/article/details/89316097 link
The above installation process is tested and effective by the author. If you have any questions, you can discuss and exchange in the comment area below.