AprilTag Detection Operator

Authors: Holoscan Team (NVIDIA)
Supported platforms: x86_64, aarch64
Language: C++, Python
Last modified: August 5, 2025
Latest version: 1.0
Minimum Holoscan SDK version: 2.2.0
Tested Holoscan SDK versions: 2.2.0
Contribution metric: Level 1 - Highly Reliable

The apriltag_detection extension provides real-time detection of April tags from different tag families. The detection and processing is performed efficiently using CUDA acceleration for optimal performance in real-time applications.

Overview

April tags are 2D barcode-like patterns that can be used for camera calibration, pose estimation, and object tracking. This operator integrates the AprilTag detection library with NVIDIA Holoscan, providing GPU-accelerated tag detection capabilities.

Features

• CUDA-accelerated processing: Leverages GPU computing for high-performance tag detection
• Multiple tag family support: Compatible with various AprilTag families
• Real-time performance: Optimized for low-latency applications

Usage

from holoscan.operators import ApriltagDetectorOp

# Create the operator
apriltag_op = ApriltagDetectorOp(
    width=1920,
    height=1080,
    number_of_tags=10
)

Parameters

Parameter	Type	Default	Description
width	int	None	Width of the input video stream in pixels. Must match the actual input resolution.
height	int	None	Height of the input video stream in pixels. Must match the actual input resolution.
number_of_tags	int	None	Maximum number of April tags to detect and output. Higher values may impact performance.

Input/Output Ports

• input: Video stream input
• output: Detected AprilTag ID and corner coordinates