A Transformer-Based Siamese Network for Change Detection

By Wele Gedara Chaminda Bandara, Vishal M. Patel. IEEE International Geoscience and Remote Sensing Symposium, 2022

1,047 citations recorded by the source corpus.

This paper presents a transformer-based Siamese network architecture (abbreviated by ChangeFormer) for Change Detection (CD) from a pair of co-registered remote sensing images. Different from recent CD frameworks, which are based on fully convolutional networks (ConvNets), the proposed method unifies hierarchically structured transformer encoder with Multi-Layer Perception (MLP) decoder in a Siamese network architecture to efficiently render multi-scale long-range details required for accurate CD. Experiments on two CD datasets show that the proposed end-to-end trainable ChangeFormer architecture achieves better CD performance than previous counterparts. Our code and pre-trained models are available at github.com/wgcban/ChangeFormer.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/ef3b15260a610473c95662f5df2c195ac19f64d6

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

1. Ming-Hsuan Yang
2. Runsheng Xu
3. Martin Danelljan
4. Fahad Shahbaz Khan
5. Michael Felsberg

In Computer Vision and Pattern Recognition

1. Runsheng Xu
2. Bin Cui
3. Yue Zhao
4. Hongruixuan Chen
5. Ming-Hsuan Yang
6. Sheng Fang
7. Wentao Zhang
8. Zhe Li

In Media Technology

1. Hongruixuan Chen
2. Ming-Hsuan Yang
3. Sheng Fang
4. Zhe Li
5. Chengxi Han
6. Kaiyu Li
7. Naoto Yokoya
8. Jian Song

Citations in the corpus

• Changer: Feature Interaction is What You Need for Change Detection (2022, background): “ChangeFormer [2] is a pure transformer model, which is a siamese variant of SegFormer [42], and is fine-tuned in network depth.”
• ChangeMamba: Remote Sensing Change Detection With Spatiotemporal State Space Model (2024, methodology): “On the Transformer front, we compare against ChangeFormer (versions 1 to 4) [22], BIT (with ResNet-18 to ResNet-101 as the backbone for feature extraction) [20], TransUNetCD [25], and SwinSUNet [24].”
• An Empirical Study of Remote Sensing Pretraining (2022, neutral): “31 ChangeFormer [144] IMP-MiT-B2 [39] — 90.”
• Diffusion Models: A Comprehensive Survey of Methods and Applications (2022, methodology): “Comparison between Diffusion-based model DDPM-CD [67] and the other SOTA model ChangeFormer [11] on remote sensing change detection task.”
• RS-Mamba for Large Remote Sensing Image Dense Prediction (2024, methodology): “In recent years, deep learning models based on transformers [7] have been widely applied to VHR remote sensing dense prediction tasks [8], [9], [10], [11].”