Parameter-Efficient Adaptation of Geospatial Foundation Models through Embedding Deflection

2025-03-12

AI Summary

To address the challenge of low-cost adaptation of geospatial foundation models (GFMs) to multispectral satellite imagery, this paper proposes DEFLECT—an embedding deflection method. DEFLECT freezes the backbone parameters and introduces only a minimal number of learnable parameters, modeling spatial priors via geometric deflection in the embedding space while integrating spectral-aware feature enhancement. It breaks from conventional low-rank fine-tuning paradigms by uniquely coupling the intrinsic structural priors of pre-trained GFMs with geometric embedding deflection, thereby significantly improving spectral representation capability. Evaluated across three GFM backbones and five cross-domain geoscience tasks, DEFLECT achieves or surpasses state-of-the-art performance while requiring only 1/5–1/10 the number of parameters compared to mainstream adaptation methods.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Remote Sensing / Geospatial AI
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

As large-scale heterogeneous data sets become increasingly available, adapting foundation models at low cost has become a key issue. Seminal works in natural language processing, e.g. Low-Rank Adaptation (LoRA), leverage the low"intrinsic rank"of parameter updates during adaptation. In this paper, we argue that incorporating stronger inductive biases in both data and models can enhance the adaptation of Geospatial Foundation Models (GFMs), pretrained on RGB satellite images, to other types of optical satellite data. Specifically, the pretrained parameters of GFMs serve as a strong prior for the spatial structure of multispectral images. For this reason, we introduce DEFLECT (Deflecting Embeddings for Finetuning Latent representations for Earth and Climate Tasks), a novel strategy for adapting GFMs to multispectral satellite imagery with very few additional parameters. DEFLECT improves the representation capabilities of the extracted features, particularly enhancing spectral information, which is essential for geoscience and environmental-related tasks. We demonstrate the effectiveness of our method across three different GFMs and five diverse datasets, ranging from forest monitoring to marine environment segmentation. Compared to competing methods, DEFLECT achieves on-par or higher accuracy with 5-10$ imes$ fewer parameters for classification and segmentation tasks. The code will be made publicly available.

Problem

• Adapting Geospatial Foundation Models to multispectral satellite imagery efficiently.
• Enhancing spectral information in features for geoscience and environmental tasks.
• Achieving high accuracy with significantly fewer parameters for classification and segmentation.

Innovation

• DEFLECT adapts GFMs with minimal parameters
• Enhances spectral info for geoscience tasks
• Achieves high accuracy with fewer parameters