Smart Edge Image Inference
Microsoft introduced its “intelligent Edge” strategy with the focus being the use of machine learning and artificial intelligence of images: a large volume of training images would be processed in the Azure Cloud using ML/AI to produce a lightweight recognition (inference) neural net that could be deployed at the edge (in a modified drone or specialised camera that each could do image inference.)
Cloud-based image processing and machine learning is a mature technology, so the limitation to realizing this strategy was the ability to do image inference at the edge easily. Drone-based intelligent image analysis is particularly tempting because of the availability of inexpensive, remotely pilotable, “consumer” drones with high-resolution cameras that revolutionize what can be done with image and video processing, except most can’t do image inference by themselves.
Microsoft is the world’s largest software vendor and a leading cloud services provider
Business Need: Take advantage of consumer drone availability to automate diverse visual tasks, fundamentally improving the business process (speed, quality) and/or very significantly reducing costs.
By using cellular connectivity to the drone, and using the mobile edge, convenient drone control and image inference can be added easily, greatly expanding the potential footprint of this kind of application.
Edge Need: Any kind of application that responds to recognized images requires short-enough latency to work with human or automated drone control (e.g., camera tracking of recognized items, hovering).
Ease of Incorporation: The biggest constraint is the adaptation of business processes to incorporate the technology, and learning how to do cloud-based AI and Machine Learning.