Feedback Neural Network based Super-resolution of DEM for generating high fidelity features
This work addresses the lack of high-resolution DEMs for many parts of the world, which is a problem for applications in environmental modeling and disaster prediction, though it appears incremental as it adapts existing feedback neural network concepts to DEM super-resolution.
The paper tackles the problem of generating high-resolution Digital Elevation Models (DEMs) from low-resolution inputs, which are crucial for applications like water flow modeling and landslide prediction, by proposing a feedback neural network architecture that iteratively adds high-frequency details. The result is that their network, DSRFB, achieves RMSEs ranging from 0.59 to 1.27 across four datasets without needing additional data like aerial images.
High resolution Digital Elevation Models(DEMs) are an important requirement for many applications like modelling water flow, landslides, avalanches etc. Yet publicly available DEMs have low resolution for most parts of the world. Despite tremendous success in image super resolution task using deep learning solutions, there are very few works that have used these powerful systems on DEMs to generate HRDEMs. Motivated from feedback neural networks, we propose a novel neural network architecture that learns to add high frequency details iteratively to low resolution DEM, turning it into a high resolution DEM without compromising its fidelity. Our experiments confirm that without any additional modality such as aerial images(RGB), our network DSRFB achieves RMSEs of 0.59 to 1.27 across 4 different datasets.