---
license: mit
language:
- en
pretty_name: EcoWikiRS
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
dataset_info:
  features:
  - name: id
    dtype: string
  - name: EUNIS
    dtype:
      class_label:
        names:
          '0': '14'
          '1': '15'
          '2': '16'
          '3': '21'
          '4': '23'
          '5': '24'
          '6': '25'
          '7': '26'
          '8': '31'
          '9': '32'
          '10': '40'
          '11': '41'
          '12': '43'
          '13': '44'
          '14': '45'
          '15': '47'
          '16': '48'
          '17': '49'
          '18': '50'
          '19': '52'
          '20': '53'
          '21': '54'
          '22': '55'
          '23': '56'
          '24': '57'
  - name: species_list
    list: string
  - name: image
    dtype: image
  - name: species_text
    dtype: string
  splits:
  - name: train
    num_bytes: 6206448837
    num_examples: 58230
  - name: validation
    num_bytes: 618946914
    num_examples: 6172
  - name: test
    num_bytes: 2829427476
    num_examples: 27398
  download_size: 7798126881
  dataset_size: 9654823227
---
# EcoWikiRS: Learning Ecological Representations of Satellite Images from Weak Supervision with Species Observations and Wikipedia

**Authors**  
Valerie Zermatten · Javiera Castillo-Navarro · Pallavi Jain · Devis Tuia · Diego Marcos  

[![arXiv](https://img.shields.io/badge/arXiv-2503.20871-red)](https://arxiv.org/abs/2504.19742)
[![GitHub](https://img.shields.io/badge/GitHub-Code-black)](https://github.com/eceo-epfl/EcoWikiRS)
[![DOI](https://img.shields.io/badge/DOI-Zenodo-blue)](https://zenodo.org/records/15236742)

## Overview 

The **WikiRS** dataset, composed of triplets of images, species list and Wikipedia sentences :
- **91k high-resolution aerial images** (50cm, RGB bands) from the swissIMAGE product 
- crowd-sourced species observations from **2745 different species**,  collected from **GBIF**, geolocated within the footprint of the aerial image.
- **103K different sentences** describing the habitat of the observed species, extracted from the corresponding Wikipedia article.

## Data formats in this repository

### 1. Original raw data
- In `ecowikirs.zip`
- Original data structure used in the codebase and archived on Zenodo
- Intended for full reproducibility

### 2. Hugging Face processed version
- In the `data/` directory
- Automatically loaded via `load_dataset("EPFL-ECEO/EcoWikiRS")`

### HF Features
- id (string): Unique site/sample identifier
- EUNIS (ClassLabel): Habitat class label. Original codes remapped to contiguous indices to handle gaps.
- species_list (list[string]): All species recorded at the site.
- image (Image): PIL image from TIFF file (`swissimage/{id}.tif` in the raw data).
- species_text (json string): Aggregated JSON data per species (from `wiki/{species_name}.json` in the raw data), keyed by binomial_name, including keywords, habitat_section, and random_sentences.

### Notes
EUNIS labels have been remapped in the Hugging Face processed version to contiguous integers. The mapping is the following:
```{0: 14, 1: 15, 2: 16, 3: 21, 4: 23, 5: 24, 6: 25, 7: 26, 8: 31, 9: 32, 10: 40, 11: 41, 12: 43, 13: 44, 14: 45, 15: 47, 16: 48, 17: 49, 18: 50, 19: 52, 20: 53, 21: 54, 22: 55, 23: 56, 24: 57}```


## How to cite this work: 

```
@InProceedings{Zermatten_2025_WikiRS,
    author    = { Zermatten, Valerie and Castillo-Navarro, Javiera and Jain, Pallavi and Tuia, Devis and Marcos, Diego},
    title     = {EcoWikiRS: Learning Ecological Representations of Satellite Images from Weak Supervision with Species Observations and Wikipedia},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    month     = {June},
    year      = {2025},
    pages     = {00-00}
}
```


## Additional data information 

- More information on the EUNIS ecosystem type map is available on the European Environment Agency website : [**Ecosystem type map** (all classes)](https://www.eea.europa.eu/en/analysis/maps-and-charts/ecosystem-type-map-all-classes-1).

- The raw aerial images with 10cm resolution from the swissIMAGE product can be openly downloaded from the [swisstopo website ](https://www.swisstopo.admin.ch/de/orthobilder-swissimage-10-cm)


## Contacts & Information

- **Contact Email:** valerie.zermatten@epfl.ch
- **Shared by:** [ECEO Lab](https://www.epfl.ch/labs/eceo/)
- **License:** MiT License