🧠 SBAI 2025 Minicourse: Advanced TinyMLOPs Methods: Implementing Machine Learning in Embedded Systems

✍🏾Authors: Thommas Flores, Daniel Costa and Ivanovitch Silva

📅 Date: July 29, 2025

📍 Event: Brazilian Symposium on Intelligent Automation (SBAI 2025)

📌 Overview

This minicourse addresses the growing demand for machine learning deployment on low-power embedded devices, driven by the evolution of edge computing. It focuses on advanced quantization strategies—including Quantization-Aware Training (QAT) and Post-Training Quantization (PTQ)—as well as model compression techniques using the TensorFlores framework.

Developed by the Conecta2AI research group (UFRN) in collaboration with LES2 (FEUP), TensorFlores automates MLP optimization and C++ code generation for microcontrollers, enabling efficient and portable AI at the edge.

🎯 Goals

General Goal

Enable participants to design and deploy optimized machine learning models on embedded systems using TensorFlores.

Specific Objectives

• Understand the mathematical foundations of model quantization.
• Apply advanced model compression techniques (PTQ and QAT) for efficient ML on resource-constrained devices.
• Gain hands-on experience in developing and deploying optimized TinyML models.

🧑‍💻 Target Audience & Prerequisites

Target Audience

Students, educators, researchers, and practitioners in Engineering, Computer Science, or related fields interested in embedded ML.

Prerequisites

Basic knowledge of Python and C++ (Arduino IDE) is required.

🗂️ Course Outline

⏱️ Duration: 3 hours

Theoretical Part (1 hour)

1. Introduction to TinyML (20 min)

• What is TinyML and why it matters
• Use cases and industry applications

• Challenges and opportunities

  1. Post-Training Quantization (PTQ) (20 min)

• Overview and best practices

• Example applications in TinyML

  1. Quantization-Aware Training (QAT) (20 min)

• Overview and best practices

• Practical use cases

Practical Part (2 hours)

1. Setting Up TensorFlores (50 min)

• Installation and environment setup
• Workflow: training, quantization, C++ code generation

• Integration with microcontrollers

  1. Training and Deploying MLP Models (50 min)

• Guide for PTQ-based deployment

• Guide for QAT-based deployment

  1. Challenges in Real-World Projects (20 min)

• Discussion of deployment challenges

• Shared experiences and practical tips

🔧 Required Materials

• Participant laptops with Python IDE and Arduino IDE installed
• Internet access
• Projector or screen for visual presentations

📤 Installation

• Install Python v3.9.6
• Clone this repository.
• Inside the cloned repository folder, open the terminal or command prompt and run:

bash python -m venv venv

• Activate the virtual environment:

bash venv\Scripts\activate

• Install the required packages inside the environment:

bash pip install -r requirements.txt

📁 Repository Structure

plaintext 📦 SBAI2025-TensorFlores ├── 📂 code │ ├── 📂 arduino │ │ ├── model.h │ │ └── tensorflores.ino │ └── 📂 notebooks │ ├── 01_EDA.ipynb │ ├── 02_data_segregation.ipynb │ ├── 03_train_standard.ipynb │ ├── 04_post_train_quantization.ipynb │ ├── 05_train_aware_quantization.ipynb │ └── my_key.py ├── 📂 data │ └── emission_total.csv ├── 📂 docs │ ├── slides.pdf │ └── Ementa.pdf ├── 📂 figures │ └── minicurso.png ├──📜 .gitignore ├──📜 LICENSE.txt ├──📜 README.MD └──📜 requirements.txt

📚 Jupyter Notebooks

• Exploratory Data Analysis (EDA)

• Data Segregation

• Training Standard Model

• Post Training Quantization

• Quantization Aware Training

📚 Arduino Codes

• Arduino ML code

🗂️ References

1. Flores, T.K.S., Costa, D.G., & Silva, I. (2025). TensorFlores: An enhanced Python-based TinyML framework. SoftwareX, 31, 102224.→ Presents the TensorFlores framework, its modular architecture, evolving clustering-based quantization, and automatic C++ code generation for embedded systems.

2. Flores, T.K.S., Medeiros, M., Silva, M., Costa, D.G., & Silva, I. (2025). Enhanced Vector Quantization for Embedded Machine Learning: A Post-Training Approach With Incremental Clustering. IEEE Access, 13, 17440–17456.

3. Flores, T.K.S., Medeiros, M., Silva, M., Costa, D.G., & Silva, I (2025). Advancing Tiny Machine Learning Operations: Robust Model Updates in the Internet of Intelligent Vehicles. IEEE Micro, vol. 45, no. 1, pp. 76-86.

   License

This package is licensed under the MIT License - © 2023 Conect2ai.