Target detection method using the annotation tool labelImg

       Target detection, the original image labeling process is very important, its role is to label the target position of the object in the original image and each image generated xml file corresponding standard indicates the position of the target frame. However markup tools bloggers reprint articles http://blog.csdn.net/u014696921/article/details/53353896 mentioned, although easy to use, but can not be labeled more of the same goals in the same image; and marked its completion after generating the corresponding txt file only requires certain tools to aid converted to the corresponding xml file. This article describes an easy to use and capable of multi-category label and can directly generate markup tools --labelImg tool xml file and use it to make a presentation.

      This article describes the use of only doing for ubuntu system.

1, download LabelImg

Embodiment 1: URL: Click the link to open the direct download, will move to the home master labelImg-master.zip download folder decompression, to give LabelImg-master file, after opening as shown.

Option 2: Use the command git

git clone https://github.com/tzutalin/labelImg
  
  

After the download is automatically extract the good in your home directory, a file named LabelImg, inside the content above chart.

2, installation

  
  
$ sudo apt-get install pyqt4-dev-tools # 安装PyQt4
$ Sudo PIP lxml install # install lxml, if error, you can try the following statement
$ sudo apt-get install python-lxml
  
  

Then open the terminal, enter the directory to use make to compile LabelImg

  
  
cd LabelImg
make all
  
  

Complete
3, using

Used in the terminal to perform directory labelImg

python labelImg.py
  
  

Running interface below

4, shortcut keys

Ctrl + u load all the images in the directory, click on Open dir same function
Ctrl + r to change the default Notes destination directory (saved xml file address)
Ctrl + S Save
Ctrl + d to copy the current tab and rectangle
space of the current image is marked as verified
w create a rectangle
at d picture
on a picture
del delete the selected rectangle
Ctrl ++ enlarged
Ctrl-- reduction
↑ → ↓ ← keyboard arrows to move a selected rectangular frame

5, specific issues

FIG want to modify the label content categories (e.g., default dog, person, cat and the like) in two modifications predefined_classes.txt master file directory in the data folder.

When used, use ctrl + u shortcut after loading images, use ctrl + location to save xml file r shortcut key assignments generated, and then begin by category picture target rectangle marked after each label a target software automatically pop up category information for selecting the pop-up category information corresponding to the category name selected by double-clicking. When a picture in each category required labeled target Mark all, or just click on the Save button to generate the corresponding xml file location information using the shortcut key ctrl + s to save, then you can mark a picture at the start.

6, attention

Now updated on github latest version may not be available on some longer use the old computer, for unknown reasons, in case the new version is not available, consider using the first labelImg old version, provided Baidu cloud address us here. In addition, some shortcuts mentioned in older versions of four of them is not available, use the mouse to click on to use the same function.

The old version Address: http://pan.baidu.com/s/1b5qyke      Password: jkz0