ArduSorter: A Real-Time Object Classification Automation System Based on Machine Learning and Arduino

ArduSorter is a web-based open-source project that combines computer vision, machine learning, and Arduino hardware control to achieve real-time object classification and automatic sorting. Its core advantages include cross-platform compatibility (runs in browsers), privacy protection (local ML inference), and ease of use (guided setup), suitable for scenarios like education, small-scale automated production lines, and prototype verification. The project is maintained by Halonatlanticwhitecedar449, with source code hosted on GitHub.

• Original Author/Maintainer: Halonatlanticwhitecedar449
• Source Platform: GitHub
• Original Link: https://github.com/Halonatlanticwhitecedar449/ArduSorter
• Release Date: June 8, 2026 This project aims to lower the barrier to using machine learning and automation technologies, allowing beginners and makers to quickly build intelligent sorting systems, which has both educational value and practical functions.

Hardware Components

• Computing Device: Computer supporting Windows/macOS/Linux + modern browser (Chrome/Firefox/Edge)
• Visual Acquisition: USB/built-in camera
• Control Core: Arduino development board (Uno/Mega, etc.) + USB data cable

Software Architecture

• Cross-platform: Runs based on browsers, no client-side development required
• Real-time Video Processing: WebRTC technology to capture video streams
• Local ML Inference: TensorFlow.js runs classification models on the browser side
• Hardware Communication: Web Serial API connects to Arduino to send control commands

Computer Vision Classification

• Supports custom object recognition (users can train models)
• Real-time processing with low latency

Arduino Control

• Sends commands to control sorting actuators (e.g., servo motors, conveyor belts, indicator lights)

User Interface

• Guided setup process
• Real-time visual feedback (camera feed + classification results)
• Parameter adjustment (classification threshold, response speed, etc.)

Education and Training

• Interdisciplinary learning (ML, computer vision, hardware programming)
• Hands-on practice and project-based learning (STEM courses)

Small-scale Automated Production Lines

• Part classification (screws, electronic components, etc.)
• Quality control (defective product sorting)
• Recycling classification (recyclable materials)

Prototype Verification

• Proof of concept (feasibility test before industrial-grade systems)
• Algorithm testing and user feedback collection

Machine Learning Model

• May use MobileNet (lightweight CNN, balancing speed and accuracy)
• Supports user-defined model training

Browser and Hardware Communication

• Web Serial API for serial communication
• Custom protocol for command transmission

Deployment Method

• Download ZIP package, unzip and run directly, no complex environment configuration required

Future Plans

• Expand support for more object categories
• Optimize user interface
• Add support for more Arduino models and compatible development boards

Project Value

• Technology Democratization: Make complex technologies accessible to ordinary users
• Educational Significance: Provide end-to-end ML application cases
• Open Source Contribution: Combine with the Arduino ecosystem to lower the threshold for innovation
• Web Technology Expansion: Demonstrate the browser's capabilities in hardware control and ML inference