What is Azure Percept?
The Azure Percept platform has three elements – the hardware, development kit, and cloud-based development and management tools.
Microsoft is working with the ecosystem of hardware developers to publish patterns and best practices for developing edge AI hardware that can be integrated easily with Azure AI and IoT services. I expect Original Device Manufacturers (ODM) to ship Percept-ready IoT and edge devices in the near future.
The Azure Percept Development Kit (DK) is a device targeting the device builder, makers, and developers to prototype AI-enabled edge solutions that can be easily deployed on production-ready hardware designed for Percept.
Microsoft has done a great job of connecting the dots across public cloud services such as Azure IoT, Azure Cognitive Services, Azure ML, and ONNX Runtime to simplify the workflow and the lifecycle management of devices and models running at the edge.
Developers can connect the Azure Percept DK device to the cloud to ingest data, build a no-code ML model, deploy the model, and perform inference. This entire workflow is available through a no-code approach. Microsoft has created a new development environment called Percept Studio that brings all the relevant cloud services under one roof. Developers can build custom vision or conversational AI applications and deploy them to the edge through Azure IoT.
A Closer Look at Azure Percept Developer Kit
Azure Percept DK is an edge AI and IoT development kit designed for developing vision and audio AI proof of concepts.
At the heart of the development kit is the carrier board that acts as an interface between the local devices and the cloud services. The Azure Percept DK carrier board is powered by an NXP iMX8m processor and a hardware-based Trusted Platform Module (TPM). It comes with multiple USB ports, WiFi and Bluetooth for connecting sensors such as cameras and microphones. It has 16GB onboard storage and 4GB of memory. It needs a 19VDC at 3.42A power supply to function.
The most interesting aspect of Azure Percept DK is the operating system based on , a Linux distribution from Microsoft for the cloud and edge infrastructure.
The kit also has two sensor boards in the form of Azure Percept Vision and Azure Percept Audio that are connected to the carrier board.
The Azure Percept Vision device comes with two Sony IMX219 Camera sensors with 6P Lens that supports 8MP resolution at 30FPS. For accelerating the AI inference, the device has an inbuilt Intel Movidius Myriad X Vision Processing Unit (VPU). When models are deployed to Azure Percept devices, the Azure Percept Studio optimizes the models to take advantage of the VPU. Azure Percept Vision is connected to the carrier board through a USB Type C connector.
Azure Percept Audio features four far-field microphones along with programmable LEDs. It can be integrated with conversational AI models for scenarios such as virtual concierge, Point of Sale (PoS) services and quality control, and warehouse task tracking.
The Percept DK carrier board, Percept Vision and Percept Audio support the interface for production deployment.
Azure Percept Studio has pre-trained models such as people detection, vehicle detection, object detection and products-on-shelf detection that can be deployed to the devices with just a few clicks.
Azure Percept Studio – The IDE for AI@Edge
According to Microsoft, Azure Percept Studio is the single launch point for creating edge AI models and solutions. Azure Percept Studio allows developers to discover and complete guided workflows that make it easy to integrate edge AI-capable hardware and powerful Azure AI and IoT cloud services.
The workflows in Azure Percept Studio integrate many underlying Azure AI and IoT services, like Azure IoT Hub, Custom Vision, Speech Studio, and Azure ML Services, so you can use these services to create an end-to-end solution without significant pre-existing knowledge. If you are already familiar with these Azure services, you can also connect to and modify existing resources outside of the Azure Percept Studio.
Azure Percept Studio abstracts multiple services and the workflow involved in connecting them through an intuitive and straightforward user interface. It delivers an integrated developer experience to build and deploy AI models at the edge.
Azure Percept platform attempts to simplify running AI at the edge. It’s a powerful edge computing platform for prototyping computer vision and conversational AI models and deploying them in production.