A Weakly-Supervised, Multitask Deep Learning Framework for Shadow Mitigation in Remote Sensing Imagery

Scott D. Couwenhoven, Emmett J. Ientilucci, Byung H. Park, David Hughes

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

We propose a weakly-supervised, multitask framework for training a convolutional neural network to solve the problem of cloud shadow mitigation given only cloud and shadow masks as labels. The network minimizes the Wasserstein distance between shadows and their proximal sunlit neighborhoods, generating a supervisory signal directly from within the input image. We extract further utility from the shadow mask through multitask learning by introducing an auxiliary task of shadow segmentation. Our approach is advantageous since it performs mitigation in an end-to-end framework which requires only a shadowed image for inference. We apply this process to the Landsat 8 OLI SPARCS validation data set and demonstrate plausible results.

Original languageEnglish
Title of host publicationIGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages619-622
Number of pages4
ISBN (Electronic)9781665427920
DOIs
StatePublished - 2022
Event2022 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2022 - Kuala Lumpur, Malaysia
Duration: Jul 17 2022Jul 22 2022

Publication series

NameInternational Geoscience and Remote Sensing Symposium (IGARSS)
Volume2022-July

Conference

Conference2022 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2022
Country/TerritoryMalaysia
CityKuala Lumpur
Period07/17/2207/22/22

Keywords

  • cloud shadow mitigation
  • deep learning
  • multitask learning
  • satellite imagery
  • weak supervision

Fingerprint

Dive into the research topics of 'A Weakly-Supervised, Multitask Deep Learning Framework for Shadow Mitigation in Remote Sensing Imagery'. Together they form a unique fingerprint.

Cite this