Embedding AI in Smart Sensors

点击翻译按钮获取中文摘要

Intempco Inc. is a Canadian corporation located in Montreal, Québec specializing in the design and fabrication of high performance sensors for use in industrial and environmental applications. These products are distinguished by high accuracy, robustness under physical extremes, and incorporation of integrated electronics facilitating integration into data acquisition systems. However, with the advent of so-called "smart" environments, there is increasing demand for sensors that can federate directly with the cloud environment and render higher level information by virtue of embedded artificial intelligence (AI).The focus of this project is to determine whether the current state of the art in machine learning (ML), particularly in convolutional neural networks (CNN), can provide specific added value to Intempco's product line without undue cost and complexity. Here added value corresponds to interpreting the meaning of data rather than simply its value, for example a proximity detector determining the cause of an alarm. To answer this question we will target a specific case in point - object detection using off-the-shelf, low cost camera arrays. This problem is typical of a wide range of challenging interpretation tasks and will provide Intempco a realistic assessment of what embedded AI can offer to their product line.