Next day fire prediction via semantic segmentation
arxiv(2024)
摘要
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.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要