PSOC™ Edge Training - Importing and profiling a pre-trained machine learning model on PSOC™ Edge

April 9, 2026 · View on GitHub

This training introduces why and how to optimize neural network (NN) models for deployment on PSOC™ Edge devices. It highlights that techniques such as pruning, sparsity, and quantization can significantly reduce model size and memory footprint while improving computational efficiency and lowering power consumption. The session includes hands-on labs demonstrating the dramatic performance and memory benefits of enabling machine learning optimizations.

Device family

How to Use This Training

  1. Download the training content.
  2. Watch the video or review the presentation at your own pace.
  3. Follow the step-by-step instructions in the training manual during the hands-on sections.

Training level

  • E3: Advanced

Pre-requisites

  • This training does not cover the basic concepts of ModusToolbox™ and PSOC™ Edge.

Tools (see training manual for versions and installation instructions)

Hardware

Duration

  • 60 min, including the video and hands-on labs

Agenda

  1. PSOC™ Edge and ML development ecosystem
  2. Machine learning: development workflows
  3. Using DEEPCRAFT™ Model Converter to deploy a pre-trained model
  4. Lab 1: Deploy a model on PSOC™ Edge using the machine learning DEEPCRAFT™ profiler
  5. Lab 2: Performance improvement across CPUs and NPUs
  6. Lab 3: Performance improvement with internal memories

Expected Outcome

  • Understand why optimizing neural network models is essential for deployment on resource-constrained edge devices
  • Learn key optimization techniques—pruning, sparsity, and particularly quantization—and when to apply them
  • Compare performance across different processor cores and memories
  • Understand the machine learning deployment workflow, specifically for pre-trained model deployment using DEEPCRAFT™ Model Converter

Content

References and resources

History

DateVersionDescription
04/09/2026**First public release