MSRS-CD Dataset link:baidu drive： https://pan.baidu.com/s/1Bb-uIh_uRFhrZaPM3i01dQ?pwd=MSRS or google drive：https://drive.google.com/file/d/1GuM8XpWVg5raLvJ2xfj5lSMYEzoBWxRO/view?usp=sharing
The MSRS-CD dataset significantly complements existing RSCD datasets in terms of image resolution, change types, dataset size, and change dimensions, further providing a new benchmark for RSCD. This dataset comprises 841 pairs of remote sensing images captured in southern Chinese cities from 2019 to 2023, with each image sized at 1024×1024 pixels and a spatial resolution of 0.5 meters. The dataset is divided into training, validation, and testing sets in a ratio of 7:1:2. As shown in Fig. 1, the main types of changes in the dataset include new buildings, suburban expansion, vegetation changes, and road construction.
Fig. 1. The MSRS-CD dataset example, where Image T1 and Image T2 represent two remote sensing images at different times, and GT denotes the ground truth labels.
This is the result of the quantitative analysis of some of the networks on the dataset. (More networks are waiting to be updated.)
| Methods | years | input | Precision | Recall | F1-score | IoU | OA | FLOPs(G) | Param(M) | Inference time(ms) |
|---|---|---|---|---|---|---|---|---|---|---|
| FCEF | 2018 | 256 | 74.99 | 66.69 | 70.59 | 45.44 | 91.80 | 3.55 | 1.35 | 1.90 |
| BIT | 2021 | 256 | 75.73 | 70.79 | 73.18 | 57.70 | 92.34 | 8.75 | 3.04 | 11.58 |
| FCCDN | 2022 | 256 | 75.56 | 71.31 | 73.37 | 65.42 | 92.36 | 12.49 | 6.31 | 15.29 |
| Changeformer | 2022 | 256 | 72.22 | 72.94 | 72.58 | 56.96 | 91.86 | 202.79 | 41.03 | 28.78 |
| SGSLN | 2023 | 256 | 77.39 | 69.73 | 73.36 | 56.28 | 92.52 | 11.50 | 6.04 | 10.31 |
| VcT | 2023 | 256 | 76.64 | 69.72 | 73.02 | 57.50 | 92.39 | 10.64 | 3.57 | 14.68 |
| EATDer | 2023 | 256 | 66.73 | 84.17 | 74.44 | 59.29 | 91.47 | 23.46 | 6.61 | 21.18 |
| AANet | 2024 | 256 | 71.94 | 77.03 | 74.40 | 59.23 | 92.17 | 24.21 | 15.82 | 10.63 |
| DFFNet | 2025 | 1024 | 76.22 | 78.69 | 77.44 | 61.82 | 93.23 | 56.93 | 23.32 | 19.63 |
Complete qualitative analysis results link: https://pan.baidu.com/s/1Fm5oOF0M3cl1fdd9pxl6lw?pwd=MSRS
Where white, green, red, and black respectively represent true positive, false negative, false positive, and true negative.
@ARTICLE{10813409,
author={Liu, Shenbo and Zhao, Dongxue and Zhou, Yuheng and Tan, Ying and He, Huang and Zhang, Zhao and Tang, Lijun},
journal={IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing},
title={Network and Dataset for Multiscale Remote Sensing Image Change Detection},
year={2025},
volume={18},
number={},
pages={2851-2866},
doi={10.1109/JSTARS.2024.3522135}
}
@ARTICLE{10942432, author={Liu, Shenbo and Zhao, Dongxue and Zhou, Yuheng and Tan, Ying and He, Huang and Zhang, Zhao and Tang, Lijun}, journal={IEEE Transactions on Geoscience and Remote Sensing}, title={Full-Scale Change Detection Network for Remote Sensing Images Based on Deep Feature Fusion}, year={2025}, volume={63}, number={}, pages={1-13}, doi={10.1109/TGRS.2025.3555171} }