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 Concepts of TinyML Pipelines
- Overview of TinyML workflow phases
- Attributes of edge hardware
- Key considerations for pipeline architecture
Data Gathering and Preparation
- Acquiring structured and sensor-derived data
- Strategies for data labeling and augmentation
- Preparing datasets for resource-constrained settings
Model Creation for TinyML
- Choosing model architectures suited for microcontrollers
- Training processes using standard ML frameworks
- Assessing model performance metrics
Model Refinement and Compression
- Quantization methods
- Pruning and weight sharing techniques
- Striking a balance between accuracy and resource constraints
Model Translation and Packaging
- Exporting models to TensorFlow Lite
- Incorporating models into embedded toolchains
- Managing model size and memory limitations
Implementation on Microcontrollers
- Flashing models onto hardware targets
- Setting up run-time environments
- Conducting real-time inference tests
Monitoring, Testing, and Verification
- Testing approaches for deployed TinyML systems
- Debugging model behavior on physical hardware
- Validating performance under field conditions
Integrating the Comprehensive End-to-End Pipeline
- Creating automated workflows
- Versioning control for data, models, and firmware
- Managing updates and iterative improvements
Wrap-up and Future Directions
Requirements
- A solid grasp of core machine learning concepts
- Hands-on experience in embedded programming
- Comfort with Python-based data processing workflows
Intended Audience
- AI specialists
- Software engineers
- Embedded systems experts
