Scholarlink

Accurate LULC classification is crucial for sustainable asset management and for understanding how climate change and time are transforming the landscape. Accurate and efficient LULC classification requires solid datasets and dependable classification models. Since satellite data, geospatial analytic tools, and classification algorithms are becoming more widely available, it is imperative to objectively assess the performance of various combinations of satellite data and classification methods to determine the most effective way for LULC classification. This study, the red-green-blue version of the EuroSAT dataset was used to fine-tune, pre-trained networks Visual Geometry Group (VGG19) and ResNet50 for LULC Classification using deep transfer learning, as opposed to training CNNs from scratch. The final results show the best model, which is ResNet50, compared to VGG19 ResNet50 gets the highest accuracy which is 93%.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Assessment of Deep Learning Techniques for Land Use and Land Cover Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Chennu Deepthi,
  • Tharun Sai Panuganti,
  • Manne Suneetha

摘要

Accurate LULC classification is crucial for sustainable asset management and for understanding how climate change and time are transforming the landscape. Accurate and efficient LULC classification requires solid datasets and dependable classification models. Since satellite data, geospatial analytic tools, and classification algorithms are becoming more widely available, it is imperative to objectively assess the performance of various combinations of satellite data and classification methods to determine the most effective way for LULC classification. This study, the red-green-blue version of the EuroSAT dataset was used to fine-tune, pre-trained networks Visual Geometry Group (VGG19) and ResNet50 for LULC Classification using deep transfer learning, as opposed to training CNNs from scratch. The final results show the best model, which is ResNet50, compared to VGG19 ResNet50 gets the highest accuracy which is 93%.