DINOWaterNet

Water extraction in complicated scenarios.

training dataset

We have built a large-scale water mapping dataset with ultra-high resolution optical remote sensing images, which can be seen below, and it can be downloaded from BaiduDisk, using the following link:

WaterDataset
Link: https://pan.baidu.com/s/1oIvJrIIYzSgaAYuEuiFIAg passkey: 1234 

model structure

We use the iFormer-s as our backbone to build DINOWaterNet. The parameter and inference time can be tested by using the following code:

python .\cal_Param.py

performance test

We have tested the DINOWaterNet on very large-scale regions, and tested over 700GB ultra-high resolution remote sensing images, see below:

software development

Based on GDAL, wxPython, etc, we have developed a user-friendly software, which can run successfully in Windows 10/11/Server operating systems, and it can be downloaded from the following link:

DINOWaterNet-Software
Link: https://pan.baidu.com/s/1LaaIqJ7IRvsRjrfoXe4jfA passkey: 1234 

Author:  Wang Z, et al. 
Title:   Large-scale Water Mapping in Complicated Scenarios with Optical Ultra High-resolution Remote Sensing Images and the DINO Foundation Model 
Journal: Under Review. 
Date:    2026