Staff Publications

Staff Publications

  • external user (warningwarning)
  • Log in as
  • language uk
  • About

    'Staff publications' is the digital repository of Wageningen University & Research

    'Staff publications' contains references to publications authored by Wageningen University staff from 1976 onward.

    Publications authored by the staff of the Research Institutes are available from 1995 onwards.

    Full text documents are added when available. The database is updated daily and currently holds about 240,000 items, of which 72,000 in open access.

    We have a manual that explains all the features 

Record number 536533
Title Land cover mapping at very high resolution with rotation equivariant CNNs : Towards small yet accurate models
Author(s) Marcos, Diego; Volpi, Michele; Kellenberger, Benjamin; Tuia, Devis
Source ISPRS Journal of Photogrammetry and Remote Sensing (2018). - ISSN 0924-2716
DOI http://dx.doi.org/10.1016/j.isprsjprs.2018.01.021
Department(s) Laboratory of Geo-information Science and Remote Sensing
PE&RC
Publication type Refereed Article in a scientific journal
Publication year 2018
Keyword(s) Deep learning - Rotation invariance - Semantic labeling - Sub-decimeter resolution
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 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.
Comments
There are no comments yet. You can post the first one!
Post a comment
 
Please log in to use this service. Login as Wageningen University & Research user or guest user in upper right hand corner of this page.