EoS-FM: Can an Ensemble of Specialist Models act as a Generalist Feature Extractor?

2025-11-26

AI Summary

Current remote sensing foundation models rely on excessively large parameters and massive datasets, incurring prohibitive computational costs and substantial environmental burdens. Method: This paper proposes EoS-FM—an efficient and sustainable foundation model framework—departing from the monolithic large-model paradigm. EoS-FM integrates multiple lightweight, task-specific ConvNeXtV2 expert models, coordinated via modular parameter freezing, federated training, structured pruning, and a continual expansion mechanism. Contribution/Results: The framework preserves strong cross-task generalization while drastically reducing resource consumption: it cuts training carbon footprint by over 90% compared to mainstream large models. Moreover, EoS-FM enables multi-institutional collaborative modeling and supports interpretable analysis. It establishes a novel foundation model paradigm for remote sensing AI that is low-barrier, highly adaptable, and environmentally responsible.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Deep Neural Architectures and Foundation Models
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Sustainability and carbon-aware systems for Web, mobile, and WoT
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Recent advances in foundation models have shown great promise in domains such as natural language processing and computer vision, and similar efforts are now emerging in the Earth Observation community. These models aim to generalize across tasks with limited supervision, reducing the need for training separate models for each task. However, current strategies, which largely focus on scaling model size and dataset volume, require prohibitive computational and data resources, limiting accessibility to only a few large institutions. Moreover, this paradigm of ever-larger models stands in stark contrast with the principles of sustainable and environmentally responsible AI, as it leads to immense carbon footprints and resource inefficiency. In this work, we present a novel and efficient alternative: an Ensemble-of-Specialists framework for building Remote Sensing Foundation Models (RSFMs). Our method decomposes the training process into lightweight, task-specific ConvNeXtV2 specialists that can be frozen and reused. This modular approach offers strong advantages in efficiency, interpretability, and extensibility. Moreover, it naturally supports federated training, pruning, and continuous specialist integration, making it particularly well-suited for collaborative and resource-constrained settings. Our framework sets a new direction for building scalable and efficient RSFMs.

Problem

• Developing efficient foundation models for Earth Observation with limited resources
• Reducing computational costs and carbon footprint in remote sensing AI
• Creating modular specialist ensembles for scalable feature extraction

Innovation

• Ensemble-of-Specialists framework for Remote Sensing Foundation Models
• Lightweight task-specific ConvNeXtV2 specialists with frozen reuse
• Modular approach supporting federated training and continuous integration