KPIs to evaluate the benefits of 5G and AI-enabled services deployment on ports logistics operations

This dataset includes KPIs measured during trials conducted at the Athens Living Lab of the 5G-LOGINNOV project. The Athens Living Lab developed a set of use cases and platforms which communicate over a private 5G NSA network with different types of end devices (5G-Trucks, 5G-Cranes, 5G-IoT, 5G UEs). The living lab use cases focused on the deployment of 5G and AI-enabled services tailored to safety/security applications as well as for improving the efficiency of daily port operations i.e., reduce costs, improve the utilization of human resources and automate logistics services.

The trials context is explained in details in the project's deliverable D3.1 – Trial methodology, planning and coordination while the data structure is described in D2.2 – Data collection and evaluation procedures. Both deliverables are provided in the dataset as metadata.

This dataset includes 2 KPIs:

A-KPI 11a  -- Inference time for both cloud  and far-edge deployments (the results are per frame)

A-KPI 11b  -- Power consumption for both cloud and far-edge deployments (the results are in watts/second)

The cloud uses a desktop GPU i.e., NVIDIA RTX 3090.

The far-edge uses an embedded GPU, i.e., NVIDIA Jetson AGX Xavier.

File naming convention

Example: RTX_3090_Inference_yolov5l_height_2160_width_3840.csv

RTX_3090 is the GPU exploited at the cloud

Inference is the metric

yolov5l is the ML object detection model used (https://github.com/ultralytics/yolov5)

Height of the image is 2160 and width 3840, i.e., 4K resolution