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
- An introduction to the stages of the TinyML process
- Key attributes of edge hardware
- Factors influencing workflow architecture
Acquisition and Processing of Data
- Gathering both structured and sensor-based data
- Methods for labeling and enhancing datasets
- Adapting data for resource-limited settings
Creating TinyML Models
- Choosing appropriate architectures for microcontrollers
- Establishing training processes with standard ML frameworks
- Assessing key metrics of model performance
Refining and Compressing Models
- Applying quantization methods
- Utilizing pruning and shared weight strategies
- Reconciling model accuracy with resource restrictions
Transforming and Packaging Models
- Exporting models for use with TensorFlow Lite
- Incorporating models into embedded development toolchains
- Addressing limitations in model size and memory
Implementation on Microcontrollers
- Loading models onto specific hardware targets
- Setting up the execution environment
- Testing real-time inference capabilities
Oversight, Verification, and Assurance
- Approaches for testing deployed TinyML systems
- Troubleshooting model behavior on physical hardware
- Validating performance under real-world conditions
Assembling the Complete Integrated Workflow
- Establishing automated processes
- Managing versions of data, models, and firmware
- Overseeing system updates and iterations
Wrap-up and Future Directions
Requirements
- A solid grasp of machine learning core concepts
- Proficiency in embedded programming
- Knowledge of data workflows utilizing Python
Intended Participants
- Artificial Intelligence engineers
- Software developers
- Specialists in embedded systems
