Thank you for sending your enquiry! One of our team members will contact you shortly.
Thank you for sending your booking! One of our team members will contact you shortly.
Duration 21 hours
Course Outline
Core Principles of TinyML Workflows
- Introduction to the phases of the TinyML lifecycle
- Attributes of edge computing hardware
- Strategic considerations for pipeline architecture
Data Acquisition and Preparation
- Gathering structured data and sensor inputs
- Methods for data labeling and augmentation
- Adapting datasets for resource-limited environments
Model Creation for TinyML
- Choosing appropriate model architectures for microcontrollers
- Training procedures utilizing standard ML frameworks
- Assessing key model performance metrics
Model Refinement and Compression
- Application of quantization methods
- Implementing pruning and weight sharing
- Achieving a balance between accuracy and resource usage
Model Transformation and Packaging
- Exporting models to TensorFlow Lite
- Incorporating models into embedded development toolchains
- Addressing model size and memory limitations
Deployment on Microcontrollers
- Writing models to hardware targets
- Setting up runtime environments
- Conducting real-time inference evaluations
Validation, Testing, and Oversight
- Approaches for testing deployed TinyML systems
- Diagnosing model behavior on physical hardware
- Verifying performance under field conditions
Assembling the Complete End-to-End Workflow
- Establishing automated pipelines
- Version control for data, models, and firmware
- Oversight of updates and iterative improvements
Review and Future Directions
Requirements
- A solid grasp of machine learning principles
- Hands-on experience with embedded programming
- Proficiency in Python-based data pipelines
Target Audience
- AI engineers
- Software developers
- Embedded systems specialists
