학술논문
Edge AI - Use Case on Google Coral Dev Board Mini
Document Type
Conference
Source
2022 IET International Conference on Engineering Technologies and Applications (IET-ICETA) Engineering Technologies and Applications (IET-ICETA), 2022 IET International Conference on. :1-2 Oct, 2022
Subject
Language
Abstract
The Google Coral Dev Board Mini leads to a great opportunity in developing applications, using object detection or image classification on the edge by having an affordable price at the same time. The evaluation board provides easy testing of AI on the edge implementations using an Edge TPU coprocessor with overall 4 TOPS. By using performance benchmarks to compare the Dev Board with the Dev Board Mini, clear differences could be identified. The Dev Board Mini turns out to have a two to fifteen times slower inference times than the Dev Board compared on the same object detection and image classification models. When running a custom trained face mask detection model using the EfficientDet-liteO architecture the Dev Board Mini still achieves 6 fps, suitable for low cost Edge AI applications.