arXiv subjects

Eva Gmelich Meijling

Publications and source records attributed to Eva Gmelich Meijling.

3 recordsLinked to original sources

arXiv2026

Now We Know? A Systematic Comparison of TerraMind and THOR

Benchmarks for Geospatial Foundation Models (GFMs) increasingly rank models by aggregate score, but such rankings obscure why models differ: how much of the gap is architecture, how much is decoder capacity, and how much is a use-case-specific artefact? This study addresses that gap through a controlled comparison of two GFMs developed under European Space Agency's $Φ$-lab with contrasting design philosophies: THOR, which introduces a compute-adaptive architecture supporting variable patch sizes and unifies Sentinel-1, -2, and -3 data at their native resolutions; and TerraMind, a multimodal generative GFM pretrained with a dual-scale token/pixel objective that enables any-to-any cross-modal generation (Thinking-in-Modalities) to infer missing sensors at inference time. Rather than reporting a single leaderboard, we investigate the axes along which the two architectures actually differ - patch size, decoder complexity, finetuning regime, input modality, and model scale - across ten use cases spanning segmentation and regression in diverse domains, including climate disaster response, methane leak detection, snow monitoring, or sea ice mapping. We find that architectural design choices - patch size and decoder type in particular - explain more performance variance than model identity itself, that the two models embody complementary investment strategies (pretraining-time scale for TerraMind versus inference-time tokenisation for THOR), and that correctly interpreting results requires dataset-level characterisation. The resulting picture is not a single winner but a set of hypotheses and a diagnostic ablation methodology that we expect to generalise to future GFMs beyond THOR and TerraMind.

cs.LG↗

arXiv2026

COP-GEN: Latent Diffusion Transformer for Copernicus Earth Observation Data

Earth observation applications increasingly rely on data from multiple sensors, including optical, radar, elevation, and land-cover. Relationships between modalities are fundamental for data integration but are inherently non-injective: identical conditioning information can correspond to multiple physically plausible observations, and should be parametrised as conditional distributions. Deterministic models, by contrast, collapse toward conditional means and fail to represent the uncertainty and variability required for tasks such as data completion and cross-sensor translation. We introduce COP-GEN, a multimodal latent diffusion transformer that models the joint distribution of heterogeneous EO modalities at their native spatial resolutions. By parameterising cross-modal mappings as conditional distributions, COP-GEN enables flexible any-to-any conditional generation, including zero-shot modality translation without task-specific retraining. Experiments show that COP-GEN generates diverse yet physically consistent realisations while maintaining strong peak fidelity across optical, radar, and elevation modalities. Qualitative and quantitative analyses demonstrate that the model captures meaningful cross-modal structure and adapts its output uncertainty as conditioning information increases. We release a stochastic benchmark built from multi-temporal Sentinel-2 observations that enables distribution-level comparison of generative EO models. On this benchmark, COP-GEN covers 90% of the real observation manifold and 63% of its per-band reflectance range, while the strongest competing method collapses to 2.8% and 18%, respectively. These results highlight the importance of stochastic generative modeling for EO and motivate evaluation protocols beyond single-reference, pointwise metrics. Website: https://miquel-espinosa.github.io/cop-gen

cs.CV↗

arXiv2025

Supervised and self-supervised land-cover segmentation & classification of the Biesbosch wetlands

Accurate wetland land-cover classification is essential for environmental monitoring, biodiversity assessment, and sustainable ecosystem management. However, the scarcity of annotated data, especially for high-resolution satellite imagery, poses a significant challenge for supervised learning approaches. To tackle this issue, this study presents a methodology for wetland land-cover segmentation and classification that adopts both supervised and self-supervised learning (SSL). We train a U-Net model from scratch on Sentinel-2 imagery across six wetland regions in the Netherlands, achieving a baseline model accuracy of 85.26%. Addressing the limited availability of labeled data, the results show that SSL pretraining with an autoencoder can improve accuracy, especially for the high-resolution imagery where it is more difficult to obtain labeled data, reaching an accuracy of 88.23%. Furthermore, we introduce a framework to scale manually annotated high-resolution labels to medium-resolution inputs. While the quantitative performance between resolutions is comparable, high-resolution imagery provides significantly sharper segmentation boundaries and finer spatial detail. As part of this work, we also contribute a curated Sentinel-2 dataset with Dynamic World labels, tailored for wetland classification tasks and made publicly available.

cs.CV↗