Ablation Study of Multidimensional Attention and Multiscale Convolution on LEVIR-CD Dataset

This paper presents the experimental results of an ablation study on the LEVIR-CD dataset, investigating the impact of multidimensional attention and multiscale convolution on image change detection. The study was conducted as part of a dissertation project.

The results demonstrate the effectiveness of both multidimensional attention and multiscale convolution in improving change detection accuracy. The ablation study revealed that the multidimensional attention module enhances the model's ability to capture long-range dependencies in the input images, while the multiscale convolution module allows the model to learn features at different scales.

The combination of these two techniques leads to significant improvements in change detection performance compared to baseline models. The study provides valuable insights into the design of efficient and accurate deep learning models for change detection tasks.

This work is particularly relevant to researchers and practitioners in the fields of remote sensing, computer vision, and machine learning. The findings can be applied to various real-world applications, such as monitoring urban sprawl, disaster assessment, and environmental change detection.