SSL4EO-S12: A Large-Scale Multi-Modal, Multi-Temporal Dataset for Self-Supervised Learning in Earth Observation

Abstract

Self-supervised pre-training bears potential to generate expressive representations without human annotation. Most pre-training in Earth observation (EO) are based on ImageNet or medium-size, labeled remote sensing (RS) datasets. We share an unlabeled RS dataset SSL4EO-S12 (Self-Supervised Learning for Earth Observation - Sentinel-1/2) to assemble a large-scale, global, multimodal, and multi-seasonal corpus of satellite imagery from the ESA Sentinel-1 \& -2 satellite missions. For EO applications we demonstrate SSL4EO-S12 to succeed in self-supervised pre-training for a set of methods: MoCo-v2, DINO, MAE, and data2vec. Resulting models yield downstream performance close to, or surpassing accuracy measures of supervised learning. In addition, pre-training on SSL4EO-S12 excels compared to existing datasets. We make openly available the dataset, related source code, and pre-trained models at https://github.com/zhu-xlab/SSL4EO-S12.

Accepted by IEEE Geoscience and Remote Sensing Magazine. 18 pages

Cited by in corpus (5)

• Self-supervised remote sensing feature learning: Learning Paradigms, Challenges, and Future Works
• Continuous Cross-resolution Remote Sensing Image Change Detection
• Better, Not Just More: Data-Centric Machine Learning for Earth Observation
• In-Domain Self-Supervised Learning Improves Remote Sensing Image Scene Classification
• Estimating Physical Information Consistency of Channel Data Augmentation for Remote Sensing Images