Remote Sensing Image Classification with the SEN12MS Dataset
This addresses the benchmarking problem for remote sensing researchers by enabling more comparable model evaluations, though it is incremental as it adapts an existing dataset rather than creating a fundamentally new one.
The authors tackled the lack of standardized large-scale datasets for remote sensing image classification by creating a classification-oriented version of the SEN12MS dataset, providing baseline results that show multi-spectral and multi-sensor data fusion outperforms conventional RGB imagery with concrete accuracy improvements (e.g., up to 15% higher accuracy in some configurations).
Image classification is one of the main drivers of the rapid developments in deep learning with convolutional neural networks for computer vision. So is the analogous task of scene classification in remote sensing. However, in contrast to the computer vision community that has long been using well-established, large-scale standard datasets to train and benchmark high-capacity models, the remote sensing community still largely relies on relatively small and often application-dependend datasets, thus lacking comparability. With this letter, we present a classification-oriented conversion of the SEN12MS dataset. Using that, we provide results for several baseline models based on two standard CNN architectures and different input data configurations. Our results support the benchmarking of remote sensing image classification and provide insights to the benefit of multi-spectral data and multi-sensor data fusion over conventional RGB imagery.