Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mapping of Land Use and Land Cover (LULC) using EuroSAT and Transfer Learning

Suman Kunwar, Jannatul Ferdush

TL;DR

This work addresses land-use/land-cover (LULC) mapping with limited labeled RS data by applying transfer learning to Vision Transformers (ViT) pretrained on ImageNet-21k and fine-tuning on RGB EuroSAT data. It evaluates TL performance across augmented and non-augmented RGB EuroSAT datasets, utilizing PyTorch and Google Colab with a Tesla T4, and compares ViT against VGG16 and ResNet-50. The study reports a peak accuracy of 99.19% and demonstrates that data augmentation improves generalization, with practical application by mapping the Kreis Borken region to inform conservation and urban planning. The findings underscore the efficacy of RGB-based ViT transfer learning for high-precision, scalable LULC mapping using Sentinel-2 data.

Abstract

As the global population continues to expand, the demand for natural resources increases. Unfortunately, human activities account for 23% of greenhouse gas emissions. On a positive note, remote sensing technologies have emerged as a valuable tool in managing our environment. These technologies allow us to monitor land use, plan urban areas, and drive advancements in areas such as agriculture, climate change mitigation, disaster recovery, and environmental monitoring. Recent advances in AI, computer vision, and earth observation data have enabled unprecedented accuracy in land use mapping. By using transfer learning and fine-tuning with RGB bands, we achieved an impressive 99.19% accuracy in land use analysis. Such findings can be used to inform conservation and urban planning policies.

Mapping of Land Use and Land Cover (LULC) using EuroSAT and Transfer Learning

TL;DR

This work addresses land-use/land-cover (LULC) mapping with limited labeled RS data by applying transfer learning to Vision Transformers (ViT) pretrained on ImageNet-21k and fine-tuning on RGB EuroSAT data. It evaluates TL performance across augmented and non-augmented RGB EuroSAT datasets, utilizing PyTorch and Google Colab with a Tesla T4, and compares ViT against VGG16 and ResNet-50. The study reports a peak accuracy of 99.19% and demonstrates that data augmentation improves generalization, with practical application by mapping the Kreis Borken region to inform conservation and urban planning. The findings underscore the efficacy of RGB-based ViT transfer learning for high-precision, scalable LULC mapping using Sentinel-2 data.

Abstract

As the global population continues to expand, the demand for natural resources increases. Unfortunately, human activities account for 23% of greenhouse gas emissions. On a positive note, remote sensing technologies have emerged as a valuable tool in managing our environment. These technologies allow us to monitor land use, plan urban areas, and drive advancements in areas such as agriculture, climate change mitigation, disaster recovery, and environmental monitoring. Recent advances in AI, computer vision, and earth observation data have enabled unprecedented accuracy in land use mapping. By using transfer learning and fine-tuning with RGB bands, we achieved an impressive 99.19% accuracy in land use analysis. Such findings can be used to inform conservation and urban planning policies.
Paper Structure (7 sections, 7 figures, 2 tables)

This paper contains 7 sections, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Materials and Methods
  4. Dataset
  5. Training and Evaluation
  6. Results
  7. Conclusion

Figures (7)

  • Figure 1: Mapping of LULC using satellite images
  • Figure 2: ViT model purposed by Dosovitskiy et al. (2021)
  • Figure 3: Sample example images of three different classes from the EuroSAT dataset.
  • Figure 4: Comparison of loss and accuracy of models at different epochs and iterations.
  • Figure 5: Both augmented with 20 epochs and non augmented with 20 epochs model was able to predict correct results with test data
  • ...and 2 more figures