From the course: Edge AI: Tools and Best Practices for Building AI Applications at the Edge
Join today to access over 23,400 courses taught by industry experts.
Edge AI performance analysis
From the course: Edge AI: Tools and Best Practices for Building AI Applications at the Edge
Edge AI performance analysis
- [Instructor] What kinds of analytics do we need to do for Edge AI? Let's enumerate the types of analytics. They mostly overlap with analytics that we do for regular machine learning models. We first have non-AI analytics, which is common for any kind of application. We look at CPU and memory utilization on the Edge to understand if sufficient resources are available for the models. We look at average and maximum loads to understand load patterns. We will analyze inference latency and end-to-end latency to understand if this is within expected limits for the use case. We also look at these trends over time to understand if there are significant changes. What about the AI side of analytics? We will analyze the accuracy of the models, including F1, precision, and recall. Data may need manual labeling for learning the ground truth. We will look at error rates for the model. When it comes to computer vision applications,…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.