Space2Ground 2.0: A Multi-Source Dataset and Framework for Agricultural Monitoring through Fusion of Street-Level and Satellite Imagery
The repository provides a complete workflow for transforming raw Mapillary street-level imagery into an analysis-ready, parcel-linked agricultural dataset suitable for crop monitoring, visual verification, and downstream machine learning applications.
📄 Preprint: https://arxiv.org/abs/2607.28247
📦 Dataset: https://doi.org/10.5281/zenodo.21219542
🏛️ Accepted at: 45th EARSeL Symposium, Athens, Greece, 29 September – 2 October 2026
Space2Ground 2.0 combines crowdsourced street-level imagery with parcel geometries to automatically associate roadside observations with agricultural parcels.
The processing pipeline includes:
- Retrieval of Mapillary image metadata
- Reconstruction of missing camera orientations
- Viewpoint projection
- Parcel-level annotation
- Image quality filtering
- Deep feature extraction
- PCA dimensionality reduction
- K-Means clustering
- Manual dataset refinement
The resulting dataset can be combined with Sentinel-1 and Sentinel-2 observations for multimodal agricultural monitoring.
Python ≥ 3.10
Main packages:
geopandas
shapely
pandas
numpy
matplotlib
opencv-python
scikit-learn
tensorflow
tqdm
requests
Install all dependencies:
pip install -r requirements.txtIf you use this repository, please cite:
@inproceedings{tsardanidis2025space2ground,
title={Space2Ground 2.0: A Multi-Source Dataset and Framework for Agricultural Monitoring through Fusion of Street-Level and Satellite Imagery},
author={Tsardanidis, Iason and Koukos, Alkiviadis and Choumos, George and Sitokonstantinou, Vasileios and Kontoes, Charalampos},
year={2025}
}This repository is released under the MIT License.
Iason Tsardanidis
Beyond Centre of EO Research & Satellite Remote Sensing
Institute for Astronomy, Astrophysics, Space Applications and Remote Sensing (IAASARS)
National Observatory of Athens (NOA)
Email: j.tsardanidis@noa.gr