Transformer backbones with mean pooling and combined self-supervised embeddings yield robust, compact representations for EO tasks that are over 500x smaller than raw data.
TerraMind-1.0-small.https : / / huggingface
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
How to Embed Matters: Evaluation of EO Embedding Design Choices
Transformer backbones with mean pooling and combined self-supervised embeddings yield robust, compact representations for EO tasks that are over 500x smaller than raw data.