spotGEO dataset
DOI10.5281/zenodo.4432143Zenodo4432143MaRDI QIDQ6696027FDOQ6696027
Dataset published at Zenodo repository.
Gurvan Lecuyer, Marcus Märtens, Mark Rutten, Dario Izzo, Daqi Liu, Dawa Derksen, Moritz von Looz, Bo Chen, Tat-Jun Chin
Publication date: 11 January 2021
Copyright license: Creative Commons Attribution 4.0 International
The spotGEO dataset is the official dataset of ESAs Kelvins competition for spotting GEO satellites. It consists of6 400 grayscaleimage sequences of the night sky. These sequences were acquired from multiple positions using a low-cost ground based telescope. Each sequence consists of 5 frames, so in total this dataset consists of 32 000 grayscale .png-images. The goal of the competition is to find geostationary objects, which often appear as very faint and smeared out blobs. The occurence of stars (streaks), clouds and other sources of noise in the image add - among other factors - to the difficulty of this task. Part of the dataset has been annotated to support machine learning approaches. Participants of the challenge are tasked to reconstruct the annotations for the remaining sequences. For a detailed description on the challenge and this dataset, visithttps://kelvins.esa.int/spot-the-geo-satellites. Version 2:This version adds the annotations for the test folder which were used to evaluatethe final score. Additionally, a small set of missing annotations thatwere discovered during the competitionfor the train part have been added.
This page was built for dataset: spotGEO dataset
