Embedding AI in Smart Sensors

Intempco Inc.是一家位于加拿大蒙特利尔的公司，专门设计和制造用于工业和环境应用的高性能传感器。这些产品的特点是高精度，在物理极端条件下的坚固性，以及集成电子设备的结合，便于集成到数据采集系统中。然而，随着所谓的“智能”环境的出现，对能够直接与云环境联合并通过嵌入式人工智能（AI）提供更高级别信息的传感器的需求越来越大。该项目的重点是确定当前机器学习（ML）的最新技术，特别是卷积神经网络（CNN），是否可以在不增加成本和复杂性的情况下为Intempco的产品线提供特定的附加价值。在这里，附加价值对应于解释数据的含义，而不仅仅是其值，例如，接近检测器确定警报的原因。为了回答这个问题，我们将针对使用现成的低成本相机阵列进行点目标检测的具体案例。这个问题是一系列具有挑战性的解释任务的典型问题，将为Intempco提供嵌入式AI可以为其产品线提供的现实评估。

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.