Convolutional Neural Networks for Land Use and Land Cover Multi-class Maps from Historical Aerial Photographs

View Full Paper

Ask AI

Bookmark

The method achieved 90% overall accuracy for predicting land use and land cover classes.
Topographic databases enabled extraction of multi-class maps without the need for manual labeling.
This approach utilized transfer learning with pretrained models from 2014 and 2020 data for improved efficiency.
The study evaluated different convolutional neural network architectures and found PSPNet with EfficientNet yielded the best results.

Discussion

No takes yet. Share an insight, caveat, or question.

Analysis reports 90% accuracy in predicting land use and land cover types from 2006 aerial images, suggesting effective automation of labeling.

Adam Kostrzewa (2025) studied this question.

View Full Paper

Ask AI

Bookmark

Analysis reports 90% accuracy in predicting land use and land cover types from 2006 aerial images, suggesting effective automation of labeling.