MPLAB ML SDK Examples

Practical examples and tutorials for working with the MPLAB ML Development Suite Python SDK.

Quick Start

1. Get an API Key

Follow the API key creation guide to set up authentication.

2. Install Dependencies

pip install mplabml pandas matplotlib seaborn

3. Run Your First Notebook

Start with getting-started.ipynb to:

• Authenticate with MPLAB ML
• Explore available SDK functions
• List your projects

Repository Structure

mplabml-sdk-examples/
├── notebooks/                               # Interactive tutorials
│   ├── getting-started.ipynb                # ✅ Start here!
│   ├── sharing-projects.ipynb               # ✅ Collaborate with others
│   ├── understanding-data.ipynb             # ✅ Prepare your data
│   ├── import-research-dataset.ipynb        # ✅ Import labeled datasets
│   ├── import-research-dataset.ipynb        # ✅ Import labeled datasets
│   ├── auto-label-with-events.ipynb         # ✅ Event-triggered labeling
│   ├── auto-label-continuous-signals.ipynb  # ✅ Label continuous signal data
│   └── auto-label-gesture-data.ipynb        # ✅ Label gesture data
│
├── datasets/                       # Sample data for examples
│   ├── series_load_sample.csv      # Arc fault data example
│   └── gesture_sample.csv          # Gesture with button events
│
├── docs/                           # Documentation
│   └── getting-started/
│       └── creating-an-api-key.md  # ✅ API key setup guide
│
└── scripts/                        # Utility scripts

Available Notebooks

✅ Ready to Use

getting-started.ipynb - Start here!

• Authentication and SDK setup
• Explore available functions
• List your projects

sharing-projects.ipynb - Collaborate with others

• Download and share complete projects
• Transfer between accounts

understanding-data.ipynb - Prepare your data

• Sequential vs wide formats
• Data quality checks and visualization
• Find and fix common issues

import-research-dataset.ipynb - Use existing labeled data

• Import from IEEE, Kaggle, or your own research
• Work with pre-labeled datasets
• Create manifest and upload to MPLAB ML

auto-label-with-events.ipynb - Automate labeling with triggers

• Use button press, GPIO, or PWM signals
• Detect rising/falling edges automatically
• Label gesture or activity segments

auto-label-continuous-signals.ipynb - Automate labeling for contiuous signal data

• For use with dataset containing continous discrete signals such as current, etc
• Expects a specific folder structure while saving data CSV
• Automatically adds labels and metadata for test and train data
• Sample dataset provided for your convenience here

auto-label-gesture-data.ipynb - Automate labeling for gesture data

• For use with gestures dataset
• Needs a trigger signal to mark the dataset with gesture information
• Trigger signal is a simple GPIO (button press) where a high indicates gesture performed
• Expects a specific folder structure while saving data CSV
• Automatically adds labels and metadata for test and train data
• Sample dataset provided for your convenience here

Using Datasets from the Repository

Datasets as CSV files are included in this repository here, for easy access in notebooks. You can load them directly:

Option 1: Clone the Repository

git clone https://github.com/MicrochipTech/mplabml-sdk-examples.git
cd mplabml-sdk-examples
jupyter notebook

Then in your notebook:

import pandas as pd

# Load arc fault sample
df = pd.read_csv('datasets/series_load_sample.csv')

# Or load gesture sample
df = pd.read_csv('datasets/gesture_sample.csv')

Option 2: Download Directly from GitHub

import pandas as pd

# Arc fault detection sample
url = 'https://raw.githubusercontent.com/MicrochipTech/mplabml-sdk-examples/main/datasets/series_load_sample.csv'
df = pd.read_csv(url)

# Gesture with button events sample
url = 'https://raw.githubusercontent.com/MicrochipTech/mplabml-sdk-examples/main/datasets/gesture_sample.csv'
df = pd.read_csv(url)

Note:

For auto-label-continuous-signals.ipynb and auto-label-gesture-data.ipynb notebooks, a large sample of dataset is collected in csv files and compressed into a zip folder (PrediciveMaintWedge.zip and MagicWandSimple.zip respectively). You can download them and then upload them to your content folder of your google colab session.

Datasets Included

Arc Fault Detection Sample

• Source: IEEE Low Voltage DC Series Arc Fault dataset
• Format: Sequential (Time, current_signal, label)
• Size: 200,000 samples (~12.5 seconds)
• Sampling rate: 16 kHz
• Labels: -1 (normal), 0 (transient), 1 (arc fault)
• Location: datasets/series_load_sample.csv

Gesture with Button Events Sample

• Format: Sequential (AccelerometerX/Y/Z, GyroscopeX/Y/Z, Trigger)
• Size: 950 samples (~9.5 seconds)
• Sampling rate: 100 Hz
• Events: 3 wave gestures with button triggers
• Location: datasets/gesture_sample.csv

Quadrature current and rpm for anomaly detection in FOC based motor control systems

• Labels: 1 (normal), 2 (unbalanced)
• Size: 10,000 samples
• Location: datasets/PrediciveMaintWedge.zip

Data Samples (zip) to autolabel gesture based applications

• Labels: 1 (horizontal), 2 (idle), 3(round), 4(vertical)
• Format: Sequential (AccelerometerX/Y/Z, GyroscopeX/Y/Z, Trigger)
• Location: datasets/MagicWandSimple.zip

Learning Paths

🚀 Complete Workflow

Follow this path to go from setup to trained model:

1. Setup → Create API Key
2. Authenticate → Getting Started
3. Prepare Data → Understanding Data Formats
4. Import Data → Choose your path:
   • Already have labels? → Import Research Dataset
   • Need to label with events? → Auto-Label with Events
5. Train Model → Use MPLAB ML web interface for pipeline building and training
6. Collaborate → Sharing Projects (optional)

📊 I Have Labeled Data

Skip straight to importing your research or proprietary datasets:

1. Create API Key - Set up authentication
2. Getting Started - Connect to MPLAB ML
3. Understanding Data - Verify your data format
4. Import Research Dataset - Upload with labels

🎯 I Need to Label Data

Gesture based demos

For gesture based applications, collect data (test and train samples) in the specified folder structure mentioned here. Ensure that you collect data with hardware triggers. Follow steps below:

1. Set up authentication → Create API Key
2. Run this notebook on colab → Auto-Label with Events

The data is labelled and metadata is added to indicate test/ train data. A project file is created and uploaded to your MPLAB ML Dev suite account.

Gesture based demos For applications where you collect samples as continuous discrete signal and the entire sample has one label, collect data (test and train samples) in the specified folder structure mentioned here. Follow steps below:

1. Set up authentication → Create API Key
2. Run this notebook on colab → Auto-Label discrete signal data The data is labelled and metadata is added to indicate test/ train data. A project file is created and uploaded to your MPLAB ML Dev suite account.

🎓 Quick Reference

Just need one thing?

• 🔑 Set up API key
• 📤 Share a project
• ✅ Check data quality
• 📥 Import labeled data
• 🔘 Auto-label gesture-based data with button/events and add test/train metadata

Requirements

• Python 3.8+
• MPLAB ML account (sign up here)
• API key (see setup guide)

Python Dependencies

mplabml
pandas
numpy
matplotlib
seaborn
jupyter

Install all at once:

pip install -r requirements.txt

What is MPLAB ML?

MPLAB ML Development Suite is Microchip's cloud-based machine learning platform for embedded systems. It enables:

• 🎯 Edge ML: Create compact supervised and anomaly-detection algorithms that can run on tiny edges for MCUs, dsPIC® DSCs and MPUs
• 📊 Data Processing: Time-series feature extraction and pipeline building
• 🔧 AutoML: Automated model training and optimization
• 💾 Knowledge Packs: Generate optimized C code for deployment
• 🚀 No ML Expertise Required: GUI-driven workflow for engineers

Contributing

This repository is growing:

• 📓 Additional example notebooks
• 📊 New sample datasets
• 📝 Documentation improvements

---

Last Updated: December 2025