EO-VAE: Towards A Multi-sensor Tokenizer for Earth Observation Data

2026-02-12

AI Summary

This work addresses the challenge of efficiently compressing and representing Earth observation data, which exhibit significant heterogeneity due to diverse sensor types and variable spectral channel configurations. To overcome this, the authors propose EO-VAE, a unified tokenizer based on a variational autoencoder architecture augmented with a dynamic hypernetwork. EO-VAE is the first model capable of flexibly encoding and reconstructing imagery with arbitrary combinations of spectral channels within a single framework, thereby eliminating the need for modality-specific training regimes. This approach substantially improves representation efficiency compared to conventional methods. Evaluated on the TerraMesh dataset, EO-VAE achieves superior reconstruction fidelity relative to the existing TerraMind tokenizer, establishing a more general and effective foundation for generative modeling in remote sensing.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Deep Generative Models & Autoencoders
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics

Abstract

State-of-the-art generative image and video models rely heavily on tokenizers that compress high-dimensional inputs into more efficient latent representations. While this paradigm has revolutionized RGB generation, Earth observation (EO) data presents unique challenges due to diverse sensor specifications and variable spectral channels. We propose EO-VAE, a multi-sensor variational autoencoder designed to serve as a foundational tokenizer for the EO domain. Unlike prior approaches that train separate tokenizers for each modality, EO-VAE utilizes a single model to encode and reconstruct flexible channel combinations via dynamic hypernetworks. Our experiments on the TerraMesh dataset demonstrate that EO-VAE achieves superior reconstruction fidelity compared to the TerraMind tokenizers, establishing a robust baseline for latent generative modeling in remote sensing.

Problem

• Earth observation
• multi-sensor
• tokenizer
• spectral channels
• latent representation

Innovation

• multi-sensor tokenizer
• variational autoencoder
• dynamic hypernetworks
• Earth observation
• latent representation