Deep learning based landcover classification algorithms have recently been proposed in literature. In hyperspectral images (HSI) they face the challenges of large dimensionality, spatial variability of spectral signatures and scarcity of labeled data. In this article we propose an end-to-end deep learning architecture that extracts band specific spectral-spatial features and performs landcover classification. The architecture has fewer independent connection weights and thus requires lesser number of training data. The method is found to outperform the highest reported accuracies on popular hyperspectral image data sets.
@article{arxiv.1612.00144,
title = {BASS Net: Band-Adaptive Spectral-Spatial Feature Learning Neural Network for Hyperspectral Image Classification},
author = {Anirban Santara and Kaustubh Mani and Pranoot Hatwar and Ankit Singh and Ankur Garg and Kirti Padia and Pabitra Mitra},
journal= {arXiv preprint arXiv:1612.00144},
year = {2017}
}
Comments
8 pages, 10 figures, Submitted to IEEE TGRS, Code available at: https://github.com/kaustubh0mani/BASS-Net