Oriented Bounding Boxes Object Detection Example

July 11, 2026 · View on GitHub

Introduction

Oriented object detection surpasses standard object detection by adding angular precision to pinpoint objects in images.

Usage

The calibration process for the rotated object detection task is as follows:

Import the image file.
Click the rotations button on the left menu bar or press the shortcut key O to quickly draw a rotated box.
Select the corresponding rotated box object and drag the rotation handle above it to adjust the angle directly, or use the following shortcut keys:

Shortcut Key	Description
z	Rotate counterclockwise by a large angle (default: 1.0°)
x	Rotate counterclockwise by a small angle (default: 0.1°)
c	Rotate clockwise by a small angle (default: 0.1°)
v	Rotate clockwise by a large angle (default: 1.0°)

To display the rotation degrees, you can click 'View' in the top menu bar, and check 'Show Degrees' to display the estimated rotation angle in real-time.

Note

Rotation Increment Customization (X-AnyLabeling v3.3.0+)

You can customize the rotation increment angles by configuring the .xanylabelingrc file in your user directory. Add or modify the following settings under the canvas section:

canvas:
  rotation:
    large_increment: 1.0  # degrees, corresponds to Z/V keys
    small_increment: 0.1  # degrees, corresponds to X/C keys

The increment values are specified in degrees and can be adjusted to suit your annotation precision requirements.