TINYML AND EMBEDDED AI WITH ESP32
Duration – 5 days.
Pre-requisites
- Basic knowledge of C/C++ programming (Arduino IDE familiarity preferred)
- Understanding of microcontrollers (GPIO, ADC, serial communication)
- Fundamental knowledge of machine learning concepts (what training and inference mean)
- Laptop with Arduino IDE installed (Windows/Linux/macOS) and USB cable for ESP32
- ESP32 development board (ESP32/ESP32-S3 recommended), basic electronic components (LED, button, potentiometer, light sensor)
Tools & Platforms
- Hardware: ESP32 board, USB cable, LED, push button, potentiometer/light sensor, breadboard & jumper wires, laptop/PC
- Software: Arduino IDE (with ESP32 support + TFLM library), Edge Impulse Studio / TensorFlow Lite, Arduino Serial Monitor/Plotter, Python (optional)
Take away
TinyML & ESP32 Workshop Outcomes
- Deploy TensorFlow Lite Micro models on ESP32 boards
- Perform basic signal processing (filtering, feature extraction) for sensor data
- Implement real-time ML applications (gesture, anomaly, keyword, and chatbot demos)
- Evaluate and optimize TinyML models for inference speed and memory footprint
- Prototype IoT + AI solutions for domains like predictive maintenance, smart monitoring, and human–device interaction
- Confidently design and demonstrate a working TinyML project using ESP32
Day 2 – ESP32 Hardware & Environment Setup
Hands-On: ESP32 Setup & Basics
- ESP32 hardware overview
- WiFi / Bluetooth
- ADC, GPIO
- Low-power modes
- Arduino IDE setup
- ESP32 board support installation
- Install TensorFlow Lite for Microcontrollers library
- Blink + Serial Print sketch (environment verification)
Day 3 – TinyML Demos on ESP32
Hands-On: TinyML Model Deployment
- Hello World TinyML model on ESP32
- Sine wave regression demo from TFLM examples
- Visualizing inference results using Arduino Serial Plotter
Day 4 – Sensor Data & Simple ML Applications
Hands-On: ESP32 Sensor Data & ML
- ESP32 peripherals: ADC, GPIO, PWM
- Data acquisition methods (serial logging, CSV exports)
- Signal processing basics
- Noise filtering
- Feature extraction (mean, RMS, FFT concepts)
- ML algorithms suitable for ESP32
- Small Neural Networks
- Decision Trees / Threshold-based ML
- Button Press Classifier using GPIO
- Analog signal anomaly detection using ADC
Day 5 – Embedded AI Applications & Mini Project
Hands-On: TinyML Applications on ESP32
- Edge ML application patterns on ESP32
- Chatbot-style TinyML (intent classification using text input)
- Predictive maintenance (sensor-based – conceptual)
- Smart energy monitoring (conceptual)
- Performance tuning on ESP32 (memory & inference constraints)
- Future of TinyML on MCUs
- Simple Chatbot implementation using Serial Monitor
- Intent-based responses
- LED control use case
