PLANTEUR: A Crop-Type Segmentation Benchmark for Tropical Regions
What it is
PLANTEUR is a benchmark for crop-type mapping from satellite imagery in the French overseas territories, built on the PASTIS design. It provides georeferenced, aligned Sentinel-2, Sentinel-1 and Landsat-8 sequences with parcel-derived labels across six tropical regions and an 8-class label space, plus a standard test split and a leave-one-island-out protocol for cross-territory generalization. The authors report evaluating fully supervised models, linear probes and fine-tuned foundation models, with code and data released on GitHub and Hugging Face.
The problem it solves
Crop-type mapping benchmarks are concentrated in temperate regions, so there is no standard way to measure how segmentation models generalize across tropical territories and sensor combinations.
What you could use it for
- Benchmarking crop-type segmentation models on tropical, multi-sensor satellite time series
- Measuring cross-territory generalization via the leave-one-island-out protocol
- Evaluating linear probes and fine-tuned foundation models against fully supervised baselines
- Training or fine-tuning remote-sensing models on aligned Sentinel-2/Sentinel-1/Landsat-8 parcel-labeled data
Why it matters
If the benchmark is adopted, it gives the remote-sensing community a standardized, multi-region tropical evaluation target where one previously did not exist, making cross-territory generalization claims comparable and exposing how well foundation models transfer outside temperate training domains.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.