Dokumentdetails
ID
oai:arXiv.org:2403.13545
Thema
Computer Science - Computer Vision... Computer Science - Machine Learnin...Autor
Alexis, Konstantinos Girtsou, Stella Apostolakis, Alexis Giannopoulos, Giorgos Kontoes, CharalamposKategorie
Computer Science
Jahr
2024
Auflistungsdatum
27.03.2024
Schlüsselwörter
day task learning computer prediction fireZusammenfassung
In this paper we present a deep learning pipeline for next day fire prediction.
The next day fire prediction task consists in learning models that receive as input the available information for an area up until a certain day, in order to predict the occurrence of fire for the next day.
Starting from our previous problem formulation as a binary classification task on instances (daily snapshots of each area) represented by tabular feature vectors, we reformulate the problem as a semantic segmentation task on images; there, each pixel corresponds to a daily snapshot of an area, while its channels represent the formerly tabular training features.
We demonstrate that this problem formulation, built within a thorough pipeline achieves state of the art results.
;Comment: Accepted in MACLEAN@ECML/PKDD 2023
Alexis, Konstantinos,Girtsou, Stella,Apostolakis, Alexis,Giannopoulos, Giorgos,Kontoes, Charalampos, 2024, Next day fire prediction via semantic segmentation
