Oct 15, 2018
Qualcomm products mentioned within this post are offered by Qualcomm Technologies, Inc. and/or its subsidiaries.
Are you ready to create IoT solutions like home monitoring cameras, enterprise security cameras and smart home devices with built-in artificial intelligence (AI)? The hardware, software and cloud components are now in place for you to get started.
A few months ago we wrote that Qualcomm Technologies and Microsoft were working together to create a vision AI developer kit that you could apply in areas like retail, manufacturing, logistics and smart home. The kit, shown below, is a camera-based device that combines IoT, edge and AI from Microsoft with Qualcomm Vision Intelligence 300 Platform and the Qualcomm Neural Processing SDK for AI for on-device edge computing.
This vision AI developer kit, a reference design for the Vision Intelligence Platform, is available for ordering now.
Hardware-accelerated vision AI at the intelligent edge
IoT connects devices; AI at the edge makes them smart.
The platform is powered by the Qualcomm Artificial Intelligence (AI) Engine so you can run machine learning models in IoT devices. Imagine a camera at your front door that is not only connected to the network but is also running a machine learning model you’ve trained. It can recognize a family member who is locked out, then connect with the door’s smart lock to allow entry.
Or, in an enterprise-related use case, imagine adding photos of a new employee to a database and pushing them to the intelligent edge. On-device vision AI can grant or deny the employee access to buildings anywhere in the world.
Instead of sending the image to the cloud for inference, waiting for processing and hoping that the network will send the response quickly, the platform can evaluate the image locally in real time. The advantages of that kind of edge computing include better security and privacy, faster processing and less need for network bandwidth.
Inside the Qualcomm AI Engine is the Qualcomm Neural Processing SDK for AI, a software framework with tools for analyzing, optimizing and debugging machine language projects. Dedicated hardware in the Qualcomm Vision Intelligence Platform, built on the Qualcomm QCS603 system on chip (SoC), accelerates the Qualcomm Neural Processing SDK.
Together, the hardware, software and cloud components are designed to deliver real-time vision AI at the edge, without the need for high-powered machines or a continuous network connection.
Getting AI to the edge
First, you train your models in Microsoft Azure for tasks like visual defect detection, object recognition, object classification and motion detection. Then, you deploy the trained models to the developer kit, where they can run even when not connected to the cloud. As shown below, Microsoft Azure takes you all the way from the cloud to the edge.
In the cloud, Qualcomm Technologies and Microsoft have integrated the Qualcomm Neural Processing SDK for AI with the Azure Machine Learning. The integration allows you to convert the AI models that you’ve used TensorFlow, Caffe/Caffe2 and the ONNX standard to pre-train or retrain.
The Azure service then packages your models into hardware acceleration-ready containers. Azure IoT Edge, a fully managed edge platform for connecting and managing IoT assets at the edge, deploys containers to run on the Qualcomm Vision Intelligence Platform. Once the models reach the platform, they take advantage of the Qualcomm Neural Processing SDK and dedicated hardware acceleration, whether on the CPU, GPU or DSP. The result is high-performance vision AI inferencing on the edge.
And, you can take things one step further. Because Azure IoT Edge runs various cloud workloads, you can deploy supported Azure services like Azure Cognitive Services, Azure Stream Analytics, Azure Functions and Azure SQL Server databases for edge analytics and AI processing.
Next step: Order the vision AI developer kit
The vision AI developer kit is a reference implementation designed to jumpstart your development of smart cameras at the network edge. It’s a solid example of how the Qualcomm Vision Intelligence Platform is being developed for use in a variety of applications including VR/360 cameras, enterprise/industrial IP cameras and consumer IP cameras.
Order your vision AI developer kit now and find out how easily you can combine the Azure Machine Learning and Azure IoT Edge services from Microsoft and the edge computing power of the Qualcomm Vision Intelligence Platform from Qualcomm Technologies.