LabelImg

If using Anaconda on Windows, use the Anaconda Prompt to install dependencies and run the build commands.

conda install pyqt=5
conda install -c anaconda lxml
pyrcc5 -o libs/resources.py resources.qrc
python labelImg.py
# Or with arguments
python labelImg.py [IMAGE_PATH] [PRE-DEFINED CLASS FILE]

To build a binary for macOS High Sierra, run the macOS build script located in the build-tools directory.

cd build-tools
./build-for-macos.sh

PASCAL VOC is the default format where annotations are saved as XML files.

1. Launch LabelImg.
2. Go to Menu/File -> Change default saved annotation folder to select your output directory.
3. Click Open Dir to select your image folder.
4. Click Create RectBox (or press w).
5. Click and drag the left mouse button to draw the bounding box.
6. Use the right mouse button to drag an existing box to move or copy it.

Annotations are saved to the folder specified in step 2.

On Windows, you need to install Python, PyQt5, and lxml manually. After installing, navigate to the labelImg directory and compile the resources.

Standard Windows Installation:

1. Run pyrcc5 -o libs/resources.py resources.qrc to compile resources.
2. Run python labelImg.py.

Windows with Anaconda:

1. Open Anaconda Prompt and navigate to the labelImg directory.
2. Install dependencies using conda install pyqt=5 and conda install -c anaconda lxml.
3. Run pyrcc5 -o libs/resources.py resources.qrc.
4. Run python labelImg.py.

# Windows + Anaconda
conda install pyqt=5
conda install -c anaconda lxml
pyrcc5 -o libs/resources.py resources.qrc
python labelImg.py

To build a Windows binary, use the containerized build process within the build-tools directory.

cd build-tools
sh run-in-container.sh
sh envsetup.sh
sh build-windows-binary.sh

To use YOLO format, follow these steps:

1. Define your classes in data/predefined_classes.txt.
2. Launch LabelImg.
3. In the toolbar, click the PascalVOC button to switch the mode to YOLO.
4. Use Open/OpenDIR to load images.
5. When you save an image, a .txt file with the same name is created in the image folder, along with a classes.txt file defining the class names.

Important Notes for YOLO:

• Do not change your class list in the middle of processing a directory; saving an image updates classes.txt, but previous annotations won't be updated to match.
• The "default class" function is not used when saving in YOLO format.
• The "difficult" flag is discarded in YOLO format.

To build labelImg from source on Ubuntu using Python 3 and Qt5, follow these steps:

sudo apt-get install pyqt5-dev-tools
sudo pip3 install -r requirements/requirements-linux-python3.txt
make qt5py3
python3 labelImg.py
# Or with arguments
python3 labelImg.py [IMAGE_PATH] [PRE-DEFINED CLASS FILE]

You can use a pre-built Docker image that contains all required dependencies.

docker run -it \
--user $(id -u) \
-e DISPLAY=unix$DISPLAY \
--workdir=$(pwd) \
--volume="/home/$USER:/home/$USER" \
--volume="/etc/group:/etc/group:ro" \
--volume="/etc/passwd:/etc/passwd:ro" \
--volume="/etc/shadow:/etc/shadow:ro" \
-v /tmp/.X11-unix:/tmp/.X11-unix \
tzutalin/py2qt4

# After container starts:
make qt4py2;./labelImg.py

You can run LabelImg in a Docker container. This requires passing the display environment and mounting necessary system volumes to allow GUI access and user permissions.

Run the following command to start the container and compile the resources for qt4py2:

docker run -it \
--user $(id -u) \
-e DISPLAY=unix$DISPLAY \
--workdir=$(pwd) \
--volume="/home/$USER:/home/$USER" \
--volume="/etc/group:/etc/group:ro" \
--volume="/etc/passwd:/etc/passwd:ro" \
--volume="/etc/shadow:/etc/shadow:ro" \
--volume="/etc/sudoers.d:/etc/sudoers.d:ro" \
-v /tmp/.X11-unix:/tmp/.X11-unix \
tzutalin/py2qt4

make qt4py2;./labelImg.py

To build LabelImg from source on Ubuntu, you need Python 3 and Qt5. Follow these steps:

1. Install pyqt5-dev-tools via apt-get.
2. Install dependencies from the provided requirements file.
3. Compile the resources using make qt5py3.
4. Run the application using python3 labelImg.py.

sudo apt-get install pyqt5-dev-tools
sudo pip3 install -r requirements/requirements-linux-python3.txt
make qt5py3
python3 labelImg.py
# Or with arguments
python3 labelImg.py [IMAGE_PATH] [PRE-DEFINED CLASS FILE]

The simplest way to install LabelImg on modern Linux distributions (like Ubuntu or Fedora) is using pip3. You can launch the application directly or provide an image path and a pre-defined class file as arguments.

pip3 install labelImg
labelImg
# Or with arguments:
labelImg [IMAGE_PATH] [PRE-DEFINED CLASS FILE]