Titlebook: Visual Inference for IoT Systems: A Practical Approach; Delia Velasco-Montero,Jorge Fernández-Berni,Angel Book 2022 The Editor(s) (if app

推测

Introduction,scale visual recognition challenge [.]. Prior to that milestone, vision algorithms were ad hoc pieces of engineering demanding painstaking efforts from senior practitioners to achieve moderate performance in real-world scenarios. AlexNet proved that highly accurate visual pipelines were possible. Si

intolerance

似少年

,Optimal Selection of Software and Models for Visual Inference,arget application specifications. For instance, power consumption is crucial for edge smart cameras, whereas real-time processing may be a requisite for other applications. This chapter describes two graphical methodologies to select the optimum combination of convolutional neural network and softwa

通便

Relevant Hardware Metrics for Performance Evaluation,l how the resources in the processor are exploited during visual inference. On the other hand, several performance parameters—throughput, power consumption, memory footprint, or CPU utilization—describe the actual system performance. More specifically, both CNN architectures—with diverse complexity

辞职

Prediction of Visual Inference Performance,sen for a specific edge application is trained, its actual implementation on a particular edge platform may not fulfill prescribed application requirements. Therefore, selecting or designing a CNN able to deliver acceptable execution time or energy consumption on the embedded system is crucial—even

Estimable

变色龙

floaters

Embedded Vision for the Internet of Things: A Survey on State-of-the-Art Hardware, Software, and Deectures currently available. A global overview of this ecosystem is depicted in Fig. .. Each application scenario is intimately related to a particular dataset to train a CNN architecture, which will be deployed on a framework supported by underlying libraries that interact with the hardware through the operating system.

induct

–LOUS

Visual Inference for IoT Systems: A Practical Approach978-3-030-90903-1