Home

Awesome

Fusing Multi-modal Data for Supervised Change Detection

Paper preview

Exemplary observations utilized in our multi-modal bi-temporal change detection work. Left: Rows of optical Sentinel-2 data as well as co-registered radar Sentinel-1 data. Each column denotes a different time point, pre- and post change. Right: The associated change map. The Sentinel-2 observations and pixel-wise annotations are by Daudt et al (2018), we curated and provide additional Sentinel-1 SAR data.


This repository presents the multi-modal change detection SAR Sentinel-1 data (available here, paper here) utilized in the work of

Ebel, Patrick und Saha, Sudipan und Zhu, Xiao Xiang (2021) Fusing Multi-modal Data for Supervised Change Detection. ISPRS. XXIV ISPRS Congress 2021, 04 - 10 July 2021, Nice, France / Virtual. paper here

Please also consider the work of

Daudt, R. C., Le Saux, B., & Boulch, A. (2018, October). Fully convolutional siamese networks for change detection. In 2018 25th IEEE International Conference on Image Processing (ICIP) (pp. 4063-4067). IEEE.

whose original data set we built upon and extended. The original Sentinel-2 observations and pixel-wise annotations are available here. We curated and provide additional Sentinel-1 SAR measurements, co-registered and temporally aligned with the original data in order to provide an opportunity for multi-sensor change detection. You can find the complementary Sentinel-1 SAR data here.

Updates:

  1. Provided code under \code. Credits: The code builds on Rodrigo's repository and extends it.

  2. You may also be interested in our follow-up publications

S. Saha, P. Ebel and X. X. Zhu, "Self-Supervised Multisensor Change Detection," in IEEE Transactions on Geoscience and Remote Sensing, doi: 10.1109/TGRS.2021.3109957. url, code

and

S. Saha, M. Shahzad, P. Ebel and X. X. Zhu, "Supervised Change Detection Using Pre-Change Optical-SAR and Post-Change SAR Data," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, doi: 10.1109/JSTARS.2022.3206898. url, code

which build on this data set, released with our ISPRS Congress publication.