Relating CNN-Transformer Fusion Network for Change Detection

TL;DR

This paper introduces RCTNet, a novel deep learning model combining CNNs and transformers with specialized modules to improve remote sensing change detection by capturing both global context and fine details.

Contribution

The paper proposes RCTNet, a CNN-transformer fusion network with novel modules for early feature fusion, enhanced temporal and multi-scale feature learning, and efficient global attention, addressing limitations of existing methods.

Findings

RCTNet outperforms traditional methods in accuracy.

It achieves a good balance between accuracy and computational efficiency.

Extensive experiments validate its superiority in remote sensing change detection.

Abstract

While deep learning, particularly convolutional neural networks (CNNs), has revolutionized remote sensing (RS) change detection (CD), existing approaches often miss crucial features due to neglecting global context and incomplete change learning. Additionally, transformer networks struggle with low-level details. RCTNet addresses these limitations by introducing \textbf{(1)} an early fusion backbone to exploit both spatial and temporal features early on, \textbf{(2)} a Cross-Stage Aggregation (CSA) module for enhanced temporal representation, \textbf{(3)} a Multi-Scale Feature Fusion (MSF) module for enriched feature extraction in the decoder, and \textbf{(4)} an Efficient Self-deciphering Attention (ESA) module utilizing transformers to capture global information and fine-grained details for accurate change detection. Extensive experiments demonstrate RCTNet's clear superiority over traditional RS image CD methods, showing significant improvement and an optimal balance between accuracy and computational cost.