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

TL;DR

EO-VAE introduces a versatile multi-sensor tokenizer for Earth observation data, enabling efficient encoding and reconstruction across diverse sensor modalities, thus advancing remote sensing generative modeling.

Contribution

It presents EO-VAE, a unified variational autoencoder that encodes multiple sensor channels with dynamic hypernetworks, unlike prior models requiring separate tokenizers.

Findings

EO-VAE outperforms TerraMind in reconstruction fidelity.

The model effectively encodes diverse sensor modalities.

Establishes a baseline for latent generative modeling in remote sensing.

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.