A new multi-source remote sensing image sample dataset with high resolution for flood area extraction: GF-FloodNet

Deep learning algorithms show good prospects for remote sensing flood monitoring. They mostly rely on huge amounts of labeled data. However, there is a lack available data in actual needs. In this paper, we propose high-resolution multi-source dataset area extraction: GF-FloodNet. GF-FloodNet contains 13388 samples from Gaofen-3 (GF-3) and Gaofen-2 (GF-2) images. We use multi-level sample selection interactive annotation strategy based active to construct it. Compare with other flood-related datasets, not only has spatial resolution up 1.5 m provides pixel-level labels, but also consists thoroughly validate evaluate the using several deep models, including quantitative analysis, qualitative validation large-scale real scenes. Experimental results reveal that significant advantages by It can support different models training extract areas. There should be potential optimal boundary model any dataset. The seems close 4824 provide at https://www.kaggle.com/datasets/pengliuair/gf-floodnet https://pan.baidu.com/s/11yx5ERsGkkfUQXPYn34KkQ?pwd=yh47.