Fully reversible neural networks for large-scale surface and sub-surface characterization via remote sensing
This addresses memory issues for researchers and practitioners in remote sensing, offering an incremental improvement over existing reversible network techniques.
The paper tackles the memory limitations of convolutional neural networks when processing large-scale hyperspectral and airborne geophysical data by using fully reversible networks, which enable training deep networks on entire data volumes for semantic segmentation without patch-based methods, as demonstrated in land-use change detection and aquifer mapping.
The large spatial/frequency scale of hyperspectral and airborne magnetic and gravitational data causes memory issues when using convolutional neural networks for (sub-) surface characterization. Recently developed fully reversible networks can mostly avoid memory limitations by virtue of having a low and fixed memory requirement for storing network states, as opposed to the typical linear memory growth with depth. Fully reversible networks enable the training of deep neural networks that take in entire data volumes, and create semantic segmentations in one go. This approach avoids the need to work in small patches or map a data patch to the class of just the central pixel. The cross-entropy loss function requires small modifications to work in conjunction with a fully reversible network and learn from sparsely sampled labels without ever seeing fully labeled ground truth. We show examples from land-use change detection from hyperspectral time-lapse data, and regional aquifer mapping from airborne geophysical and geological data.