2 min

Microsoft has made its Vision AI Developer Kit widely available through Arrow Electronics. The developer kit is a hardware base built on Qualcomm’s Vision Intelligence Platform and designed to run artificial intelligence (AI) models locally and integrate with Azure ML and Azure IoT Edge.

The development of the Vision AI Developer Kit was already announced at the Build Conference in May 2018. Then Microsoft said it was working with Qualcomm on a kit that works within Azure IoT Edge.

The kit is now ready and for sale. For example, the Vision AI Developer Kit can be used to create apps that ensure that everyone on a construction site wears a helmet, says Microsoft’s principal project manager Anne Yang to Venturebeat. Another use case is an app that detects which products are no longer present in the box in a shop.

AI workloads contain megabytes of data and possibly billions of calculations. It is now possible in hardware to run time-sensitive AI workloads at the edge, while outputs are also sent to the cloud for downstream applications.

Hardware Specifications

The Vision AI Developer Kit runs on Yocto Linux and contains a Qualcomm Snapdragon 603. This combined with 4GB of LDDR4X and 64GB of storage. There is also an 8 megapixel camera sensor that can film in 4K UHD. Four microphones handle sounds and commands.

The whole makes connection via wifi, and then specifically via 802.11b/g/n 2.4Ghz/5Ghz. There is also an HDMI out port, as well as audio in and out ports and a USB-C port. The internal storage can be expanded with a microSD card.

The associated containerised Azure services are performed using the Snapdragon Neural Processing Engine (SNPE) within Qualcomm’s Vision Intelligence 300 Platform.

Use of developer kit

The developer kit is for sale through Arrow Electronics for 249 dollars (226 euros). There is also a software development kit available via GitHub, containing Vision Studio Code with Python modules, a pre-built Azure IoT deployment configurations and a Vision AI Developer Kit extension for Visual Studio.

Developers working with Azure ML can use Azure ML to create and monitor AI models. Azure IoT Edge can be used to manage and deploy models.

Developers can create a vision model by uploading tagged images to Azure Blob Storage. Azure Custom Vision Service can then do the rest. Another option is to use Jupyter notebooks and Visual Studio Code to create your own vision models via Azure Machine Learning (AML). The trained models then collect to be converted to DLC format and packed into an IoT Edge module.