Enhancing Aerial Scene Classification through Deep Metric Learning and Ensemble Methods
Deep metric learning approaches can outperform traditional pre-trained convolutional neural networks for aerial scene classification tasks. Combining diverse deep metric learning classifiers through the Univariate Marginal Distribution Algorithm (UMDA) can further improve classification performance.
Enhanced ResNet with Convolutional Block Attention Module for Accurate Ship Classification on Optical Satellite Imagery
This study proposes an Enhanced ResNet model that integrates the Convolutional Block Attention Module (CBAM) to achieve high accuracy in classifying ships from optical satellite imagery, outperforming traditional methods.
Automated High-Precision Lake Extraction via Two-Stage Prompt-Enhanced Segmentation
A novel two-stage prompt enhancement framework, LEPrompter, leverages prompt-based training and prompt-free inference to significantly improve the accuracy of automated lake extraction from remote sensing imagery.
Comprehensive Change Interpretation and Analysis: Integrating Pixel-level Detection and Semantic-level Captioning
The proposed Change-Agent can simultaneously achieve precise pixel-level change detection and semantic-level change captioning, providing comprehensive and interactive interpretation of surface changes.
SA-MixNet: Structure-aware Mixup and Invariance Learning for Scribble-supervised Road Extraction in Remote Sensing Images
The author argues that the degradation in performance of weakly supervised road extractors is due to poor model invariance to scenes with different complexities. To address this, they propose SA-MixNet, a data-driven framework that enhances model invariance and improves road extraction performance.
© 2024 by Linnk AI