A LSTM-UNet and Zero Padding technique to detect deforestation in Amazon area

Irham Muhammad Fadhil, Aniati Murni Arymurthy

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

2 Citations (Scopus)

Abstract

The Amazon Rainforest is the largest forest in the world that stores various kinds of biodiversity, both flora and fauna. The protection of the integrity and sustainability of this rainforest is a concern for the entire international community. One form of protection is by mapping the deforestation areas by using deep learning. This paper proposes a novel Deep Learning method that combines U-Net with LSTM and Zero Padding in each convolution layer in U-net to map deforestation areas. Boundary between deforested and non-deforested area is made to boost the overall precision of the model. Generally, the proposed method indicates good accuracy in mapping the deforestation areas, which is 93.35% with an F1-score of 93.82% and a low loss value of 0.1654, while boundary use slightly boosted the overall precision into 94.06% because the use of boundaries aims to limit areas with very narrow class differences.

Original languageEnglish
Title of host publicationProceedings - 2022 IEEE International Conference on Cybernetics and Computational Intelligence, CyberneticsCom 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages218-223
Number of pages6
ISBN (Electronic)9781665497428
DOIs
Publication statusPublished - 2022
Event6th IEEE International Conference on Cybernetics and Computational Intelligence, CyberneticsCom 2022 - Virtual, Malang, Indonesia
Duration: 16 Jun 202218 Jun 2022

Publication series

NameProceedings - 2022 IEEE International Conference on Cybernetics and Computational Intelligence, CyberneticsCom 2022

Conference

Conference6th IEEE International Conference on Cybernetics and Computational Intelligence, CyberneticsCom 2022
Country/TerritoryIndonesia
CityVirtual, Malang
Period16/06/2218/06/22

Keywords

  • deforestation
  • LSTM
  • neural network
  • U-Net
  • Zero Padding

Fingerprint

Dive into the research topics of 'A LSTM-UNet and Zero Padding technique to detect deforestation in Amazon area'. Together they form a unique fingerprint.

Cite this