mlsanityData to 14 Oct 2026

Self-supervised pre-training enhances change detection in Sentinel-2 imagery

arXiv 2101.08122 ↗First posted 20 Jan 2021cs.CV eess.IVcited by 21

Not on the map. Why?

Abstract (arXiv)

While annotated images for change detection using satellite imagery are scarce and costly to obtain, there is a wealth of unlabeled images being generated every day. In order to leverage these data to learn an image representation more adequate for change detection, we explore methods that exploit the temporal consistency of Sentinel-2 times series to obtain a usable self-supervised learning signal. For this, we build and make publicly available (https://zenodo.org/record/4280482) the Sentinel-2 Multitemporal Cities Pairs (S2MTCP) dataset, containing multitemporal image pairs from 1520 urban areas worldwide. We test the results of multiple self-supervised learning methods for pre-training models for change detection and apply it on a public change detection dataset made of Sentinel-2 image pairs (OSCD).

Key facts. Cited by 21 papers in the map (data through 14 Oct 2026). Builds on 10 of its 32 references that are in the map. First posted on arXiv 20 Jan 2021.

Builds on

10 of 32 references.

References · 10

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2018Unsupervised Representation Learning by Predicting Image Rotationscited by 1,591
  3. 2014How transferable are features in deep neural networks?cited by 1,225
  4. 2015Unsupervised Visual Representation Learning by Context Predictioncited by 1,066
  5. 2018Fully Convolutional Siamese Networks for Change Detectioncited by 306
  6. 2019Revisiting Self-Supervised Visual Representation Learningcited by 251
  7. 2017Deep learning in remote sensing: a reviewcited by 128
  8. 2018Urban Change Detection for Multispectral Earth Observation Using Convolutional Neural Networkscited by 110
  9. 2018Learning Spectral-Spatial-Temporal Features via a Recurrent Convolutional Neural Network for Change Detection in Multispectral Imagerycited by 53
  10. 2019Guided Anisotropic Diffusion and Iterative Learning for Weakly Supervised Change Detectioncited by 10

Built on by

Cited by year

Citing papers per year: 2021: 4; 2022: 9; 2023: 3; 2024: 1; 2025: 4.

21 papers.

Sort:

  1. 2022Self-supervised Learning in Remote Sensing: A Reviewcited by 137
  2. 2023SSL4EO-L: Datasets and Foundation Models for Landsat Imagerycited by 49
  3. 2022EarthNets: Empowering AI in Earth Observationcited by 40
  4. 2021Change Detection Meets Visual Question Answeringcited by 29
  5. 2021Self-supervised Audiovisual Representation Learning for Remote Sensing Datacited by 24
  6. 2022Self-supervised remote sensing feature learning: Learning Paradigms, Challenges, and Future Workscited by 24
  7. 2022Semantic-aware Dense Representation Learning for Remote Sensing Image Change Detectioncited by 16
  8. 2022Inferring 3D change detection from bitemporal optical imagescited by 10
  9. 2024Generalizable Disaster Damage Assessment via Change Detection with Vision Foundation Modelcited by 10
  10. 2021Self-supervised Remote Sensing Images Change Detection at Pixel-levelcited by 9
  11. 2025SMARTIES: Spectrum-Aware Multi-Sensor Auto-Encoder for Remote Sensing Imagescited by 9
  12. 2021Evaluating Self and Semi-Supervised Methods for Remote Sensing Segmentation Taskscited by 7
9 more
  1. 2025Be the Change You Want to See: Revisiting Remote Sensing Change Detection Practicescited by 6
  2. 2022Semantic decoupled representation learning for remote sensing image change detectioncited by 4
  3. 2022Supervising Remote Sensing Change Detection Models with 3D Surface Semanticscited by 4
  4. 2022Self-supervised learning -- A way to minimize time and effort for precision agriculture?cited by 4
  5. 2023Deep Metric Learning for Unsupervised Remote Sensing Change Detectioncited by 4
  6. 2023Unsupervised CD in satellite image time series by contrastive learning and feature trackingcited by 3
  7. 2025EuroMineNet: A Multitemporal Sentinel-2 Benchmark for Spatiotemporal Mining Footprint Analysis in the European Union (2015-2024)cited by 1
  8. 2022Contrastive pretraining for semantic segmentation is robust to noisy positive pairscited by 0
  9. 2025Geodiffussr: Generative Terrain Texturing with Elevation Fidelitycited by 0