README.md

---
pretty_name: FranceCrops
license: cc-by-sa-4.0
tags:
- earth-observation
- remote-sensing
- agriculture
- timeseries
- geospatial
task_categories:
- image-classification
size_categories:
- 10K<n<100K
---

# Dataset Card for FranceCrops

## Dataset Summary

FranceCrops is a satellite imagery time series dataset for crop classification in France. 

Each sample in the dataset consists of:
- **Input Features (`x`)**: 3D arrays of shape (100, 60, 12) in float16:
  - 100: number of timeseries sampled within an agricultural field
  - 60: temporal dimension (measurements every 5 days from 01/02/2022 to 30/11/2022)
  - 12: spectral bands from Sentinel-2 satellite
- **Labels (`y`)**: Integer class labels (int16), one of 20 crop types

## Splits

🚧 Work in Progress 🚧

- `train`: 20,000 samples (~3.4GB)

## Usage

```python
from datasets import load_dataset
# Load the dataset
dataset = load_dataset("saget-antoine/francecrops", split="train")

# Example of accessing a single sample
sample = dataset[0]
x = sample["x"]  # Get the input features
y = sample["y"]  # Get the label
```

Typical usage with PyTorch:

```python
import torch
from torch.utils.data import DataLoader
from datasets import load_dataset

dataset = load_dataset("saget-antoine/francecrops", split="train").with_format("torch", columns=["x", "y"], dtype=torch.float16)

train_loader = DataLoader(dataset, batch_size=1024, shuffle=True, num_workers=8)
```

## Dataset Creation

### Source Data

🚧 Work in Progress 🚧

- Features: Sentinel-2 L2A satellite imagery extracted using GEE
- Labels: Crop type classification from the 2022French Registre Parcellaire Graphique (RPG)

### Preprocessing

🚧 Work in Progress 🚧

- Remove clouds, shadows, and missing data time steps
- Temporal alignment
- Temporal interpolation when missing

## License

Creative Commons Attribution-ShareAlike 4.0 International (CC-BY-SA-4.0)