This repository provides a suite of applications for the Kendryte K210 AI chip, targeting drone autonomy. It enables real-time object detection, classification, and marker recognition, integrating with ArduPilot flight controllers to enhance drone intelligence for tasks like visual navigation and target tracking, particularly in GPS-denied environments.

How It Works

The project leverages the K210's edge AI capabilities, employing both traditional machine vision techniques (via the OpenMV library) for tasks like color and marker detection, and deep learning models (TensorFlow, YOLOv2) for more complex recognition like faces and human bodies. It establishes a MAVLink communication interface to transmit processed target location data to ArduPilot, facilitating real-time flight control adjustments.

Quick Start & Requirements

• Installation: Requires flashing custom firmware (MaxiPy) to the K210 board using kflash_gui.
• Prerequisites: Kendryte K210 evaluation board (e.g., Sipeed MAXI DOCK K210), MicroPython environment, TensorFlow/YOLOv2 models.
• Resources: Detailed user manual (English/Chinese) available.

Highlighted Details

• Implements face, human body, and mask recognition using deep learning.
• Supports traditional vision tasks: largest color block, QR code, and AprilTag recognition.
• Provides MAVLink communication for ArduPilot integration.
• Designed as a supporting project for the Vision_MAV autonomous drone system.

Maintenance & Community

• Developed by Beihang University's Institute of Unmanned System, led by Associate Professor Li Dawei.
• Project is intended for future open-sourcing after confidentiality periods.

Licensing & Compatibility

• License details are not explicitly stated in the provided README.
• Compatibility with commercial or closed-source projects is not specified.

Limitations & Caveats

The repository is a supporting component of a larger research project (Vision_MAV) and may not be a standalone, fully polished solution. Specific hardware dependencies (K210 evaluation board) are required.