changeformer

ChangeFormer

A Transformer-Based Siamese Network for Change Detection

Introduction

Abstract

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 is available at this https URL

@misc{bandara2022transformerbased,
      title={A Transformer-Based Siamese Network for Change Detection}, 
      author={Wele Gedara Chaminda Bandara and Vishal M. Patel},
      year={2022},
      eprint={2201.01293},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Results and models

LEVIR-CD

Method	Backbone	Crop Size	Lr schd	Mem (GB)	Precision	Recall	F1-Score	IoU	config	download
ChangeFormer	MIT-B0	256x256	40000	-	93.00	88.17	90.52	82.68	config
ChangeFormer	MIT-B1	256x256	40000	-	92.59	89.68	91.11	83.67	config

• All metrics are based on the category "change".
• All scores are computed on the test set.
• We simply convert the Segformer to a siamese variant and do not strictly refer to the ChangeFormer.