Edge-X: Cost Factor Evaluation Workflow for Model Deployment on Edge AI Devices

Edge Artificial intelligence (Edge AI) is a crucial part of many new applications, including event detection, gesture recognition, on-sensor process control, and more effective and secure data transfer between systems. There is an increasing number of hardware choices for the deployment of AI models while designing such applications. It is important to understand the compute costs to effectively select a target hardware board to host such models. In this paper, we propose Edge-X, a cost factor evaluation workflow for the effective selection of hardware to host such models given input parameters. Edge-X prescribes a step-by-step guide for launching an application on an Edge AI device. The workflow is demonstrated using a proof of concept gesture recognition application deployed on three target hard-ware platforms serving as examples of complex, medium, and simple AI integrations on edge devices. The designed workflow shows capability to select appropriate edge devices quickly for any class of a desired application with AI capability.