Critical Infrastructures – CI dataset

CI dataset (Critical Infrastructures) is a dataset prepared for the detection task with Deep Learning models.

The critical infrastructure datasets that are provided in this webpage are focused on the development of an automatic critical infrastructure detection system through satellite images.

The automatic critical infrastructure detection system analyses a satellite image through a detector and returns the infrastructures that appear in the image. This system is achieved through Deep Learning (DL) techniques based on convolutional neural networks (CNN) that can be trained to detect this type of objects.
The critical infrastructure detection task depends on the input image since the satellite images are given according to the zoom level. This zoom level indicates the actual size in meters occupied by a pixel of the image. As each infrastructure can be detected in a range of zoom levels, it is possible to obtain specialized detectors in ranges of zoom levels. The strategy to perform the detections can be to apply all detectors to an image, or to apply the DetDSCI methodology to classify the input image according to its zoom level to select its specialized detector. However, any DL model requires to learn a quality image dataset and annotation according to the detection task.

Critical Infrastructure – CI dataset provides quality image datasets built for training DL models in the framework of developing an automatic critical infrastructure detection system. Each dataset for the image object detection task is described below and can be downloaded.

Download:

The public datasets are organized depending on the objects included in the dataset and the range of zoom levels. More detail about the database and access to the repository: https://github.com/ari-dasci/OD-CriticalInfrastructures

Publication date:

May 2021

Contact:

Francisco Pérez Hernández

Scroll Up