ThinkingEarth: Exploring self-supervised learning techniques

News Item
30 August 2024

Self-supervised learning on large-scale satellite data has gained interest for creating Earth observation (EO) foundation models
Self-supervised learning on large-scale satellite data has gained interest for creating Earth observation (EO) foundation models

However, valuable resources like global land-use data and powerful vision models are often overlooked. The publication “Multi-label Guided Soft Contrastive Learning for Efficient Earth Observation Pretraining “, co-authored by ThinkingEarth partner Xiao Xiang Zhu, shows that using these free resources can overcome common issues in contrastive learning and improve the efficiency of EO pretraining.

Specifically, they introduce soft contrastive learning, which uses multi-label supervision from land-cover data to better handle multiple positive matches in complex scenes. They also explore cross-domain continual pretraining using multispectral and SAR imagery, leveraging top vision models like DINOv2. By incorporating simple techniques like weight initialisation and Siamese masking, they achieve strong pretraining results, even with different input types and channels.

A lot of research has focused on developing Earth observation (EO) foundation models, mainly by adapting self-supervised pretraining techniques for EO data. Early studies focused on creating contrastive views from EO data.

What is ThinkingEarth doing with self-supervised learning

We’ll focus on improving self-supervised learning for Copernicus Sentinel datasets, especially for contrastive learning and masked image modelling in Sentinel-1, 2, and 3 data compression. Our approach will develop methods to combine self-supervised representations from different Sentinel data types, keeping important information while considering which data type is most dominant.

We also aim to boost the ability of deep learning models to generalise across different locations and times using self-supervised learning, meta-learning, and continual learning techniques.

Share

Read next


ThinkingEarth’s role in the Causal Impact of Humanitarian Aid on Food Security
News Item
30 July 2024

ThinkingEarth’s role in the Causal Impact of Humanitarian Aid on Food Security

As climate change causes more droughts, vulnerable regions face serious food shortages, increasing the need for humanitarian aid. Communities that rely on rainfall for crop growth are especially at risk and prone to economic loss, risk of malnutrition and famine and subsequent devastating loss.
ThinkingEarth for Solar Energy Applications
News Item
29 June 2024

ThinkingEarth for Solar Energy Applications

New and innovative ideas on how to harness the power of solar energy are consistently cropping up. One such challenge however is the optimal planning of photovoltaic (PV) installations on existing building rooftops.
Unleashing the power of Earth Foundation Models with ThinkingEarth
News Item
31 May 2024

Unleashing the power of Earth Foundation Models with ThinkingEarth

While there has been continuous earth foundation model developments in recent years, there are still large gaps in existing benchmarks. ThinkingEarth aims to address these gaps while also harnessing the power of Artificial Intelligence and Copernicus Foundation Models.
Newsletter of the project Thinking Earth

Stay tuned and subscribe to our quarterly newsletter

By submitting your e-mail address, you agree to our privacy policy for the site.