Détection dense de changements par réseaux de neurones siamois

Published in Conference Reconnaissance des Formes, Images, Apprentissage et Perception, RFIAP, 2018

Authors: R. Caye Daudt, B. Le Saux, A. Boulch, Y. Gousseau

Resources

Link to HAL repository Download paper here

Abstract

This paper presents convolutional neural network architectures which perform change detection using a pair of coregistered images. Most notably, we propose Siamese extensions of fully convolutional networks which use heuristics about the current problem to achieve the best re- sults in our tests on two open change detection datasets, using both RGB and multispectral images. We show that our system is able to learn from scratch using annotated change detection images. Our architectures achieve better performance than previously proposed methods, while being at least 500 times faster than related systems. We also present a change detection dataset that was developed using Sentinel-2 images.

Citation

@inproceedings{boulch:hal-01823684,
  TITLE = { {D{\'e}tection dense de changements par r{\'e}seaux de neurones siamois} },
  AUTHOR = {Boulch, Alexandre and Daudt, Rodrigo Caye and Le Saux, Bertrand and Gousseau, Yann},
  URL = {https://hal.archives-ouvertes.fr/hal-01823684},
  BOOKTITLE = { {Reconnaissance des Formes, Image, Apprentissage et Perception (RFIAP)} },
  ADDRESS = {Marne-la-Vall{\'e}e, France},
  YEAR = {2018},
  MONTH = Jun,
  KEYWORDS = {Change detection ; machine learning ; fully convolutional networks ; Earth observation},
  PDF = {https://hal.archives-ouvertes.fr/hal-01823684/file/2018_rfiap_changement.pdf},
  HAL_ID = {hal-01823684},
  HAL_VERSION = {v1},
}