Improving Remote Sensing Change Detection Via Locality Induction on Feed-forward Vision Transformer
Dublin Core
Title
Improving Remote Sensing Change Detection Via Locality Induction on Feed-forward Vision Transformer
Subject
Change Detection, Vision Transformer, Pyramidal Vision Transformer, Local Vision Transformer, CDD, LEVIR-CD
Description
The main objective of Change Detection (CD) is to gather change information from bi-temporal remote sensing images. The recent development of the CD method uses the recently proposed Vision Transformer (ViT) backbone. Despite ViT being superior to Convolutional Neural Networks (CNN) at modeling long-range dependencies, ViT lacks a locality mechanism, a critical property of pixels that comprise natural images, including remote sensing images. This issue leads to segmentation artifacts such as imperfect changed region boundaries on the predicted change map. To address this problem, we propose LocalCD, a novel CD method that imposes the locality mechanism into the Transformer encoder. It replaces the Transformer’s feed-forward network using an efficient depth-wise convolution between two 1 × 1 convolutions. LocalCD outperforms ChangeFormer by a significant margin. Specifically, it achieves an F1-score of 0.9548 and 0.9243 on CDD and LEVIR-CD datasets.
Creator
Lhuqita Fazry, Mgs M Luthfi Ramadhan, Wisnu Jatmiko
Source
http://dx.doi.org/10.21609/jiki.v17i1.1188
Publisher
Faculty of Computer Science Universitas Indonesia
Date
2024-02-25
Contributor
Sri Wahyuni
Rights
e-ISSN : 2502-9274 printed ISSN : 2088-7051
Format
PDF
Language
English
Type
Text
Coverage
Jurnal Ilmu Komputer dan Informasi (Journal of Computer Science and Information)
Files
Collection
Citation
Lhuqita Fazry, Mgs M Luthfi Ramadhan, Wisnu Jatmiko, “Improving Remote Sensing Change Detection Via Locality Induction on Feed-forward Vision Transformer,” Repository Horizon University Indonesia, accessed May 22, 2025, https://repository.horizon.ac.id/items/show/8865.