Embedded Machine Learning (Embedded ML) Training with Raspberry Pi
Duration – 5 Days
Objectives
- Learn basics of Embedded ML
- Set up Raspberry Pi for projects
- Collect sensor data and train ML models
- Run ML models on Raspberry Pi
- Use camera for real-time object detection
Tools & Platforms
- Hardware: Raspberry Pi, DHT11, MCP3008, LDR, LEDs, Pi Camera
- Software: Python, OpenCV, TensorFlow Lite, Raspberry Pi OS
- Platforms: Raspberry Pi, Jupyter Notebook
Pre-requisites
- Basic Python knowledge
- Simple electronics understanding (sensors, GPIO)
- Laptop or monitor setup for Pi
Take away
- Build real-time smart ML projects
- Collect and analyze sensor data
- Train and deploy ML models on Raspberry Pi
- Work with computer vision using camera feed
- Control GPIO based on ML predictions.
Day 2: Sensor Interfacing & Data Collection
- DHT11 Sensor: Temperature & Humidity
- Reading analog values using ADC (MCP3008 + LDR)
- Storing sensor data in CSV format
- Exploratory Data Analysis (EDA) using Python.
Hands-On: - Interfacing DHT11 - Logging sensor data into CSV - Plotting sensor data using matplotlib
Day 3: Introduction to Machine Learning Concepts
- Supervised vs Unsupervised Learning
- Regression and Classification
- real-world examples
- Training a temperature classifier
- Converting models to TensorFlow Lite format
- Model accuracy and evaluationDC/DC Converter & Battery Modeling
Hands-On: - Train an SVM/Decision Tree model in Python - Export to. tflite using TensorFlow - Test model accuracy on unseen data
Day 4: Inverter, Control Systems & PID Optimization
- Running a TensorFlow Lite model on Raspberry Pi
- Performing inference with live sensor data
- Optimizing model size and latency.
Hands-On: - Load. tflite model in Python - Predict temperature category (e.g., “Hot”, “Normal”, “Cold”) - Trigger GPIO actions based on prediction (e.g., turn on fan LED)
Day 5: Embedded Computer Vision
- Basics of Computer Vision
- Capturing video feed using Pi camera or USB webcam
- Image pre-processing and feature extraction
- Real-time object classification using MobileNet + TensorFlow Lite.
Hands-On: - Real-time object detection (e.g., identify “person” or “bottle”) - Display result on video feed - GPIO action based on object (e.g., buzz if “person” detected)
