Lightweight Attention-Enhanced DeepLabV3+ for Agricultural Land Segmentation and AI-Assisted Decision-Making Using Remote Sensing Images
Computer and Decision Making: An International Journal · 12.11.2026 · 4 (1) · 145-161
DOI: 10.1922/comdem.v4i1.19092 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 12.11.2026. Verfasst von Jayshri P. Patil und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The detection involves correctly categorizing the smishing message and Accurate agricultural land-cover segmentation can provide a spatial evidence layer for land-use assessment and agricultural planning, but image heterogeneity, class ambiguity and the cost of high-resolution inference complicate its use. This study evaluates a lightweight DeepLabV3+ segmentation configuration with an ImageNet-pretrained MobileNetV2 encoder and a post-decoder Squeeze-and-Excitation (SE) attention module. The framework was assessed on the DeepGlobe Land Cover Classification dataset using an existing crop-based evaluation protocol. The reported experiments yield an Agriculture-class Intersection over Union (IoU) of 77.68%, a six-class mean IoU (mIoU) of 69.42%, overall accuracy of 94.05%, and approximately 14 million model parameters. Relative to the reported ResNet50-based DeepLabV3+ configuration (approximately 41 million parameters), the proposed configuration has about 65.9% fewer parameters. Beyond the segmentation contribution, this study specifies a conceptual, human-supervised decision-support pathway in which predicted agricultural masks could inform land-cover summaries, spatial review and field-verification priorities. That pathway is an interpretation of the existing segmentation outputs, not an experimentally implemented or validated decision-support system. The findings support the potential relevance of a compact segmentation model to agricultural information workflows, subject to confirmation of evaluation-set reporting, uncertainty handling, georeferencing and operational decision utility.
Zitieren
APA
Patil, J. P., & Dave, D. N. (2026). Lightweight Attention-Enhanced DeepLabV3+ for Agricultural Land Segmentation and AI-Assisted Decision-Making Using Remote Sensing Images. Computer and Decision Making: An International Journal, 4(1), 145-161. https://doi.org/10.1922/comdem.v4i1.19092
Vancouver
Patil JP, Dave DN. Lightweight Attention-Enhanced DeepLabV3+ for Agricultural Land Segmentation and AI-Assisted Decision-Making Using Remote Sensing Images. Computer and Decision Making: An International Journal. 2026;4(1):145-161. doi:10.1922/comdem.v4i1.19092
BibTeX
@article{Patil2026,
author = {Patil, Jayshri P. and Dave, Dr. Namrata},
title = {Lightweight Attention-Enhanced DeepLabV3+ for Agricultural Land Segmentation and AI-Assisted Decision-Making Using Remote Sensing Images},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {4},
number = {1},
pages = {145--161},
doi = {10.1922/comdem.v4i1.19092}
}
EUVIMED