SpectralGPT: Remote Sensing Foundation Model for Spectral Data

SpectralGPT is the first purpose-built foundation model designed explicitly for spectral remote sensing data. The model considers unique characteristics of spectral data, such as spatial-spectral coupling and spectral sequentiality, within a masked autoencoder framework. The release includes trained models (SpectralGPT, SpectralGPT+), a new benchmark dataset (SegMunich) for semantic segmentation, original code, and implementation instructions.

Use Cases

• Semantic segmentation of land cover based on the SegMunich benchmark dataset.
• Pre-training foundation models for remote sensing tasks using the described 3D GPT network.
• Analyzing spatial-spectral coupling in satellite imagery using the model's architecture.
• Benchmarking new remote sensing models against the released SpectralGPT models.

Strengths

• First purpose-built foundation model for spectral remote sensing data.
• Model architecture considers spatial-spectral coupling and spectral sequentiality.
• Release includes a new benchmark dataset (SegMunich) for semantic segmentation.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Danfeng Hong, Chinese Academy of Sciences

Collection Method

Likely developed as part of research for the SpectralGPT model.