Bayesian Convolutional Neural Networks for Limited Data Hyperspectral Remote Sensing Image Classification

Hyperspectral remote sensing (HSRS) images have high dimensionality, and labeling HSRS data is expensive and therefore limited to small amounts of pixels. This makes it challenging to use deep neural networks for HSRS image classification. In extreme cases, deep neural networks are even outperformed by traditional models. In this work, we propose to use Bayesian convolutional neural networks (BCNNs) as a potential alternative to convolutional neural networks (CNNs). BCNNs benefit from Bayesian learning, which is more robust against overfitting and inherently provides a measure for uncertainty. We show in experiments on the Pavia Centre, Salinas, and Botswana datasets that a BCNN outperforms a similarly constructed non-Bayesian CNN, an off-the-shelf random forest (RF), and a state-of-the-art Bayesian neural network (BNN). We also show that BCNN is more robust against overfitting compared with the CNN. Furthermore, the BCNN exhibits a remarkably larger capacity for model compression, which makes BCNN a better candidate in hardware-constrained settings. Finally, we show that the BCNN's uncertainty measure can effectively identify misclassified samples. This useful property can be used to detect mislabeled data or to reject predictions with low confidence.

Automated summary

Hyperspectral remote sensing (HSRS) images with high dimensionality pose challenges for deep neural networks due to limited labeled data. In this study, Bayesian convolutional neural networks (BCNNs) are introduced as an alternative to CNNs, benefiting from Bayesian learning and providing an uncertainty measure. Experimental results on multiple datasets demonstrate that BCNN outperforms non-Bayesian CNNs, random forests, and Bayesian neural networks. BCNN also exhibits better resistance to overfitting and larger capacity for model compression, making it suitable for hardware-constrained settings. The uncertainty measure of BCNN effectively identifies misclassified samples, providing useful information for detecting mislabeled data or rejecting low-confidence predictions.