Microchip Technology has announced the release of the VectorBlox™ 3.0 Accelerator Software Development Kit (SDK), designed to simplify AI implementation on FPGA-based platforms while improving performance and energy efficiency for edge AI applications.
Available free of charge, the SDK provides an integrated development environment for optimizing, compiling and deploying convolutional neural network (CNN) models on PolarFire® FPGA and SoC devices. The platform enables developers to consolidate multiple vision and sensor-based AI workloads on a single low-power FPGA, helping reduce system complexity and accelerate time to market.
A key enhancement in VectorBlox 3.0 is support for sparse neural networks, allowing the accelerator to skip zero-value operations during AI inference. This approach reduces computational and memory requirements while maintaining model accuracy, enabling faster inference and lower power consumption for always-on edge AI systems.
Microchip also highlighted real-world deployments of the technology in space applications. Planetek Italia has used VectorBlox acceleration on PolarFire SoC devices to support onboard Earth observation processing, including object detection and semantic scene analysis aboard the AI-eXpress-1 satellite. The company also highlighted AIKO’s use of the platform to deploy autonomous onboard AI capabilities for cloud and ship detection in orbit.
The SDK integrates with Microchip’s Libero® SoC Design Suite and CoreVectorBlox IP, providing developers with a unified toolchain for FPGA-based AI development targeting aerospace, defense, industrial and other mission-critical applications.
Source: Microchip Technology.
