AI-Based Squirrel Detection System Using ESP32-S3

AI-Based Squirrel Detection System Using ESP32-S3

What is this project about?

I built an AI-based squirrel detection system using an ESP32-S3 CAM V1.2, Edge Impulse, and an Arduino Mega 2560. The goal was to use a camera and machine learning to identify whether a squirrel is in front of the camera and show the result using LEDs.

The system uses a red LED to indicate when a squirrel is detected and a green LED when a squirrel is not detected. This project helped me explore how artificial intelligence, computer vision, and hardware can work together in a real-world application.

Why did I decide to make it?

I wanted to try building something that combines AI with physical hardware instead of just running a machine learning model on a computer. I was interested in learning how a small camera board could use a trained model to recognize an animal and then communicate that result to another microcontroller.

I also wanted to understand the challenges involved in building a complete system, including collecting images, training a model, connecting the boards, and testing the results.

How does it work?

The ESP32-S3 camera captures images, and a machine learning model trained with Edge Impulse classifies what it sees as either squirrel or non squirrel.

When the model detects a squirrel with a confidence score that meets the configured threshold, the ESP32-S3 sends a HIGH signal through a logic-level converter to the Arduino Mega. The Mega turns on the red LED and turns off the green LED.

When a squirrel is not detected, the ESP32-S3 sends a LOW signal. The Mega then turns on the green LED and turns off the red LED.

I used a logic-level converter to connect the ESP32-S3's 3.3V signal safely to the Arduino Mega's 5V logic input.

What I learned

Building this project taught me that getting an AI model to run is only one part of creating a working system. I also had to work through wiring, power, camera capture, signal communication, and software compilation issues.

I learned how to train and test an image-classification model, connect two microcontroller boards, and use the model's output to control physical LEDs. I would like to improve the model with more varied images and test it under different lighting conditions and distances.

Images and demonstration

I have included photos of my hardware setup and wiring to show how the components are connected. I also plan to include a demonstration showing the red LED when a squirrel is detected and the green LED when one is not detected.