Infineon’s DEEPCRAFT AI Suite Unlocks Edge AI on PSOC Edge Microcontrollers
The new AI suite includes DEEPCRAFT Studio for training computer vision, audio, and radar models and DEEPCRAFT Model Converter that supports open-source frameworks such as PyTorch and TFLite for efficient deployment.
The proliferation of IoT and advanced sensor systems has intensified the demand for efficient machine learning (ML) processing directly at the edge. To address the complexities of transitioning ML models from development environments to energy-constrained embedded silicon, Infineon has introduced the DEEPCRAFT AI Suite, a comprehensive framework optimized for the company’s PSOC Edge family of microcontrollers (MCUs). This ecosystem is engineered to support a continuous workflow, from initial model development and optimization through secure deployment on hardware.
Infineon’s PSOC Edge MCUs bring high efficiency and performance to edge AI applications. Image used courtesy of Infineon Technologies
PSOC Edge MCUs
The effectiveness of the DEEPCRAFT AI Suite is tied directly to the capabilities of its target platform, the PSOC Edge MCUs. These are Arm Cortex-M-based devices characterized by their focus on high-performance compute within a low-power envelope. Crucially, the architecture includes dedicated ML acceleration hardware. This specialized silicon offloads compute-intensive neural network operations, improving overall throughput and allowing the Cortex-M cores to manage other system tasks, thereby reducing latency and power consumption.
For applications requiring local visualization, the MCUs integrate a dedicated 2.5D GPU. According to Infineon, this peripheral handles the rendering of high-quality graphical interfaces at a lower performance and energy cost than traditional 3D graphics processors.
Security is managed via a robust hardware implementation. The PSOC Edge family meets the requirements for Infineon Edge Protect Category 4 and incorporates a secure enclave, offering advanced features like PSA Certified L2 and L4 iSE and PCI pre-certification to protect intellectual property and data integrity across the device lifecycle.
The DEEPCRAFT AI Suite serves as an Edge AI ecosystem for the PSOC Edge MCUs. Image used courtesy of Infineon Technologies
Software Architecture for Model Development
The DEEPCRAFT AI Suite is not a single application, but a collection of interoperating software tools and resources designed to streamline the engineering process. All components are accessible through the centralized DEEPCRAFT AI Hub and are integrated into the existing ModusToolbox environment, which also supports development via Zephyr RTOS.
The core development platform is DEEPCRAFT Studio, an end-to-end environment for creating and training ML models optimized for the edge. While it previously supported time-series data, audio, and radar applications, the Studio now also includes comprehensive support for computer vision. This includes specific functionality for training and deploying robust object detection models using advanced YOLO models. This integration addresses the rising demand for on-device image processing in industrial and consumer sectors.
The DEEPCRAFT AI Suite comprises a bundle of tools, including DEEPCRAFT AI Hub, which functions as a “one-stop shop,” and DEEPCRAFT Studio for AI model development. Image used courtesy of Infineon Technologies
Framework Agnostic Optimization
A significant challenge in embedded AI is converting models trained in high-level frameworks for efficient execution on an MCU’s unique architecture. The DEEPCRAFT Model Converter addresses this by providing a mechanism to optimize both proprietary and open-source models for execution on PSOC Edge and PSOC 6 MCUs.
The converter is designed to be framework-agnostic, supporting standard industry tools including PyTorch, TFLite, and Keras. By simplifying the process of adapting these popular formats to the target silicon, the converter helps designers maximize the efficiency of the on-chip ML acceleration hardware.
The AI suite is rounded out by the DEEPCRAFT Voice & Audio Solutions, providing necessary toolchains for engineers developing sophisticated, high-quality, voice-controlled systems at the edge.
Edge AI Applications
The DEEPCRAFT AI Suite establishes a full-stack engineering environment that integrates software tools with the specialized hardware capabilities of the PSOC Edge MCU platform. By offering a direct path from model creation (via Studio) to hardware optimization (via Converter), the suite reduces the typical friction encountered when deploying complex ML algorithms. This systematic approach allows designers to focus on application design rather than toolchain integration. Practical applications for this highly integrated, secure, and low-power Edge AI platform span multiple sectors, including industrial automation, automotive systems, and advanced consumer electronics.
