Land cover mapping at very high resolution with rotation equivariant CNNs: towards small yet accurate models

TL;DR

This paper introduces RotEqNet, a rotation-equivariant CNN architecture for high-resolution land cover mapping that achieves high accuracy with significantly fewer parameters by encoding rotation equivariance directly into the model.

Contribution

The paper presents RotEqNet, a novel CNN design that encodes rotation equivariance using rotating convolutions and orientation vector fields, enabling small yet accurate models for remote sensing tasks.

Findings

RotEqNet outperforms standard CNNs in semantic labeling accuracy.

RotEqNet requires an order of magnitude fewer parameters.

RotEqNet achieves state-of-the-art results on sub-decimeter resolution datasets.

Abstract

In remote sensing images, the absolute orientation of objects is arbitrary. Depending on an object's orientation and on a sensor's flight path, objects of the same semantic class can be observed in different orientations in the same image. Equivariance to rotation, in this context understood as responding with a rotated semantic label map when subject to a rotation of the input image, is therefore a very desirable feature, in particular for high capacity models, such as Convolutional Neural Networks (CNNs). If rotation equivariance is encoded in the network, the model is confronted with a simpler task and does not need to learn specific (and redundant) weights to address rotated versions of the same object class. In this work we propose a CNN architecture called Rotation Equivariant Vector Field Network (RotEqNet) to encode rotation equivariance in the network itself. By using rotating convolutions as building blocks and passing only the the values corresponding to the maximally activating orientation throughout the network in the form of orientation encoding vector fields, RotEqNet treats rotated versions of the same object with the same filter bank and therefore achieves state-of-the-art performances even when using very small architectures trained from scratch. We test RotEqNet in two challenging sub-decimeter resolution semantic labeling problems, and show that we can perform better than a standard CNN while requiring one order of magnitude less parameters.