Wiznet makers

Summary: An exemplary IoT project that overcomes the hardware limitations of an 8-bit microcontroller (Arduino) using the Cloud (AWS), incorporating a voice interface (Alexa) for visually impaired users.

1. Introduction: When Technology Becomes Warm

Technology shines brightest when it solves someone's real-world discomfort. The 'Alexa Controlled Face Recognizing Arduino Door Bell' project on Hackster.io is a heartwarming invention designed for a grandmother with failing eyesight who struggles to see who is at the door.

This project goes beyond a simple doorbell that just rings; it is a smart system where the user can ask, "Alexa, who is at the door?" and the AI identifies the visitor and announces their name. The most surprising aspect is that this high-performance AI system originates from a humble, low-cost Arduino board.

2. System Architecture: The Marriage of Low-Tech and High-Tech

The defining technical characteristic of this project is the combination of "Lightweight Edge" and "Advanced Cloud."

Hardware: Simple and Cost-Effective

Instead of a high-performance processor, the setup uses very basic components to keep complexity and cost down.

• MCU: Arduino Uno (Basic 8-bit board)
• Network: Arduino Ethernet Shield 2 (Internet connectivity)
• Camera: Arducam Mini OV2640 (SPI-based low-cost camera module)
• Input: Push Button (Doorbell)

Cloud & Software: Heavy Lifting and Smart Logic

The lack of computing power on the hardware side is compensated by the powerful serverless architecture of AWS (Amazon Web Services).

• AWS API Gateway & Lambda: Acts as the entry point for Arduino requests and executes logic.
• AWS S3: Storage for captured images.
• AWS Rekognition: (Core) The AI engine that analyzes images to identify faces.
• Alexa Skills Kit: Handles the voice interaction with the user.

3. Core Technology: Crossing the 2KB Memory Barrier

The most significant technical challenge in this project is the "Memory Constraint."

The Arduino Uno (ATmega328P) has only 2KB of SRAM. In contrast, a single JPEG image captured by the camera module is tens of KB in size. It is impossible for the Arduino to store the entire image in its memory before sending it.

The Solution: Chunked Transfer (Streaming)

The development team adopted a clever method to bypass local storage:

• Buffer: The image is temporarily stored in the Arducam module's own internal buffer.
• Read: The Arduino reads data from this buffer in very small Chunks.
• Transmit: Without storing the chunk, the Arduino immediately sends it to the AWS Cloud via an HTTP PUT request.
• Repeat: This process repeats until the entire image is transmitted.

This approach opens the door for low-spec MCUs to send high-resolution images to the cloud for heavy AI processing.

4. User Experience (UX): A Doorbell That Learns

This system doesn't just recognize pre-defined people; it interacts with the user to evolve.

• Visitor Arrives: The button is pressed, and the photo is sent to AWS.
• Verification: Grandma asks, "Alexa, ask smart camera who rang the door?"
  • Known Face: "It is Peter."
  • Unknown Face: "I do not recognize the person."
• Active Learning: Grandma tells Alexa, "That person is Thomas." The system then tags the previously 'Unknown' photo with the name 'Thomas' in the database.
• Result: The next time Thomas visits, the system will recognize him.

This is a brilliant implementation of a Human-in-the-loop machine learning pipeline, applied intuitively in a home appliance.

5. Conclusion & Insights

This project offers important takeaways for IoT developers:

• Appropriate Tech: You don't always need expensive hardware; proper use of Cloud APIs enables high-level functionality.
• Accessibility-First Design: Targeting a specific user group (elderly, visually impaired) by choosing VUI (Voice) over GUI (Screen) was a key design decision.
• Scalability: Since the "brain" resides in the cloud, the software logic remains reusable even if the hardware is swapped for an ESP32 or another chipset later.

[Proposal for WIZnet Makers] Recreating the 'AI Doorbell' with ESP32-CAM

What if WIZnet upgraded this excellent project to align with modern trends?

Instead of the bulky 'Arduino Uno + Ethernet Shield' combination, I propose recreating this project using the ESP32-CAM, which is currently the most popular board among makers. By utilizing the ESP32-CAM, you can drastically reduce hardware size and cost while securing significantly faster transmission speeds.

If WIZnet creates and shares a '2025 Version AI Doorbell' reference using this board, it would serve as a highly practical and engaging tutorial for many developers in the community.