# HyperspectralMAE: The Hyperspectral Imagery Classification Model using Fourier-Encoded Dual-Branch Masked Autoencoder

arXiv: https://arxiv.org/abs/2505.05710 · HTML: https://mlsanity.com/paper/2505.05710/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 9 of its 19 references that are in the map.
- First posted on arXiv 9 May 2025.

## Builds on (most cited first, 9 of 9)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [SpectralFormer: Rethinking Hyperspectral Image Classification with Transformers](https://mlsanity.com/paper/2107.02988/) (2021)
- [Graph Convolutional Networks for Hyperspectral Image Classification](https://mlsanity.com/paper/2008.02457/) (2020)
- [Cascaded Recurrent Neural Networks for Hyperspectral Image Classification](https://mlsanity.com/paper/1902.10858/) (2019)
- [SS-MAE: Spatial-Spectral Masked Auto-Encoder for Multi-Source Remote Sensing Image Classification](https://mlsanity.com/paper/2311.04442/) (2023)
- [EnergyFormer: Energy Attention with Fourier Embedding for Hyperspectral Image Classification](https://mlsanity.com/paper/2503.08239/) (2025)
- [Locality-Aware Hyperspectral Classification](https://mlsanity.com/paper/2309.01561/) (2023)

## Built on by (most cited first, 1 of 1)

- [DisentangleFormer: Spatial-Channel Decoupling for Multi-Channel Vision](https://mlsanity.com/paper/2512.04314/) (2025)

## Abstract (arXiv)

Hyperspectral imagery provides rich spectral detail but poses unique challenges because of its high dimensionality in both spatial and spectral domains. We propose textitHyperspectralMAE, a Transformer-based foundation model for hyperspectral data that employs a textitdual masking strategy: during pre-training we randomly occlude 50% of spatial patches and 50% of spectral bands. This forces the model to learn representations capable of reconstructing missing information across both dimensions. To encode spectral order, we introduce learnable harmonic Fourier positional embeddings based on wavelength. The reconstruction objective combines mean-squared error (MSE) with the spectral angle mapper (SAM) to balance pixel-level accuracy and spectral-shape fidelity. The resulting model contains about 1.8×108 parameters and produces 768-dimensional embeddings, giving it sufficient capacity for transfer learning. We pre-trained HyperspectralMAE on two large hyperspectral corpora -- NASA EO-1 Hyperion (~1,600 scenes, ~3×1011 pixel spectra) and DLR EnMAP Level-0 (~1,300 scenes, ~3×1011 pixel spectra) -- and fine-tuned it for land-cover classification on the Indian Pines benchmark. HyperspectralMAE achieves state-of-the-art transfer-learning accuracy on Indian Pines, confirming that masked dual-dimensional pre-training yields robust spectral-spatial representations. These results demonstrate that dual masking and wavelength-aware embeddings advance hyperspectral image reconstruction and downstream analysis.