Home

ML4Earth Logo Text

The new national center of excellence ML4Earth (Machine Learning for Earth observation) brings together “hardcore” AI experts, remote sensing scientists and Earth scientists to tackle fundamental Earth observation domain-specific machine learning challenges. We demonstrate the impact of ML4Earth with a wide range of use cases related to the realization of a Digital Twin of Earth. It is also our mission to shape a strong Earth observation Data Science Community.

About ML4Earth

Our research follows a high-risk / high-gain approach. We work highly interdisciplinary and make best use of Data Science (Artificial intelligence / machine learning) for Earth observation data, having in mind grand challenges of society via our covered application fields, ultimately contributing to the Digital Twin of Earth.

Our unique selling point is our profound expertise in AI and data science in Earth observation. We pioneer the development of novel AI methods for big Earth observation data and form one of the leading research centers worldwide in that field. The research community can benefit from our expertise by attending our workshops or using benchmark datasets that we provide.

Research and
Applications

The main research focus of ML4Earth is to bring together novel AI methods for Earth observation purposes and Earth and Climate Sciences. We are covering six major research branches, each linked to a specific application field.

Physics-aware machine learning

With physics-aware machine learning, we use physical laws and physics-domain-specific knowledge to improve the performance and validity of our data-driven machine learning models.

Publications

Publications of our research team will be listed here.
Interpretable deep learning for consistent large-scale urban population estimation using Earth observation data
Interpretable deep learning for consistent large-scale urban population estimation using Earth observation data
April 1, 2024
Accurate and up-to-date mapping of the human population is fundamental for a wide range of disciplines, from effective governance and establishing policies to disaster management and crisis dilution. The traditional method of gathering population data through census is costly and time-consuming. Recently, with the availability of large amounts of Earth observation data sets, deep learning methods have been explored for population estimation; however, they are either limited by census data availability, inter-regional evaluations, or transparency. In this paper, we present an end-to-end interpretable deep learning framework for large-scale population estimation at a resolution of 1 km that uses only the publicly available data sets and does not rely on census data for inference. The architecture is based on a modification of the common ResNet-50 architecture tailored to analyze both image-like and vector-like data. Our best model outperforms the baseline random forest model by improving the RMSE by around 9% and also surpasses the community standard product, GHS-POP, thus yielding promising results. Furthermore, we improve the transparency of the proposed model by employing an explainable AI technique that identified land use information to be the most relevant feature for population estimation. We expect the improved interpretation of the model outcome will inspire both academic and non-academic end users, particularly those investigating urbanization or sub-urbanization trends, to have confidence in the deep learning methods for population estimation.

Featured paper

Accurate and up-to-date mapping of the human population is fundamental for a wide range of disciplines, from effective governance and establishing policies to disaster management and crisis dilution. The traditional method of gathering population data through census is costly and time-consuming. Recently, with the availability of large amounts of Earth observation data sets, deep learning methods have been explored for population estimation; however, they are either limited by census data availability, inter-regional evaluations, or transparency. In this paper, we present an end-to-end interpretable deep learning framework for large-scale population estimation at a resolution of 1 km that uses only the publicly available data sets and does not rely on census data for inference. The architecture is based on a modification of the common ResNet-50 architecture tailored to analyze both image-like and vector-like data. Our best model outperforms the baseline random forest model by improving the RMSE by around 9% and also surpasses the community standard product, GHS-POP, thus yielding promising results. Furthermore, we improve the transparency of the proposed model by employing an explainable AI technique that identified land use information to be the most relevant feature for population estimation. We expect the improved interpretation of the model outcome will inspire both academic and non-academic end users, particularly those investigating urbanization or sub-urbanization trends, to have confidence in the deep learning methods for population estimation.

Products and Publications

An important goal of ML4Earth is to build and maintain an international community within the AI4EO domain. We are pursuing this goal together with the Space Agency of the German Aerospace Center. We are doing so by providing benchmark datasets as a service to the community, while offering the community opportunities for training and networking.