Ayala-Romero, Jose A and Garcia-Saavedra, Andres and Costa-Perez, Xavier and Iosifidis, George
Members of: NEC
Dataset of measurements of performance and power consumption of an AI service at the network edge. The experimental setup is comprised of a 3GPP R10-compliant LTE base station (BS), a user equipment (UE) generating service requests via the BS to a well-known object recognition service, and an off-the-shelf server with an NVIDIA GPU running the service. Each request consists of an image with a variable number of objects from the COCO dataset. The images are sent to the service via the uplink channel of the LTE interface, and the service returns bounding boxes and a classification label of the identified objects to the user via the downlink channel of the LTE interface.
https://github.com/h2020daemon/SPLC22September 8, 2022