Spaces:
Running
Running
Updated org card with current project details
Browse files- codebase and website are now live, not "coming soon"
- Renamed "water body detection" to "flood detection" to match repos
- Added dataset table: scene counts, patch counts, label sources, repo links
- Added patch dimensions, Sentinel-1 IW mode, VV/VH polarisation detail
- Added geographic coverage per use case
- Add licence: MIT, with AGPL-3.0 for Ultralytics YOLO components
- Linked model repos and consortium partners
README.md
CHANGED
|
@@ -8,29 +8,55 @@ pinned: false
|
|
| 8 |
---
|
| 9 |

|
| 10 |
|
| 11 |
-
#
|
| 12 |
|
| 13 |
OpenSAR Insight is an early-phase project developing ML-ready datasets and algorithms for direct insight generation from raw SAR data. This is a first step toward efficient onboard implementation of end-to-end SAR inference pipelines for low-latency applications.
|
| 14 |
|
| 15 |
-
|
| 16 |
|
| 17 |
-
##
|
| 18 |
|
| 19 |
-
We compiled novel datasets for three use cases: vessel detection,
|
| 20 |
|
| 21 |
-
|
| 22 |
|
| 23 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
-
|
| 26 |
|
| 27 |
-
|
| 28 |
|
| 29 |
-
|
| 30 |
|
| 31 |
-
##
|
| 32 |
-
The consortium is composed of Indra Space, the University of Alcala, INTA (National Institute of Aerospace Technology), and ESA’s Φ-lab.
|
| 33 |
|
| 34 |
-
|
| 35 |
|
| 36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
---
|
| 9 |

|
| 10 |
|
| 11 |
+
# OpenSAR Insight
|
| 12 |
|
| 13 |
OpenSAR Insight is an early-phase project developing ML-ready datasets and algorithms for direct insight generation from raw SAR data. This is a first step toward efficient onboard implementation of end-to-end SAR inference pipelines for low-latency applications.
|
| 14 |
|
| 15 |
+
The project is funded and supported by ESA's Φ-lab.
|
| 16 |
|
| 17 |
+
## Datasets
|
| 18 |
|
| 19 |
+
We compiled novel datasets for three use cases: vessel detection, flood detection, and radio frequency interference (RFI) detection. All are built from Copernicus Sentinel-1 (S1A/S1B) acquisitions in Interferometric Wide (IW) mode, using both VV and VH polarisations.
|
| 20 |
|
| 21 |
+
Each dataset provides co-registered patches at multiple processing levels — Level-0 RAW, Single Look Complex (SLC), and Ground Range Detected (GRD) — with Level-0 extracted from the Level-1 products used. Labels are supplied as XML alongside the imagery, and are partially derived from SARFish/xView3, Kuro Siwo, and Aresys respectively.
|
| 22 |
|
| 23 |
+
| Dataset | Scenes | Unique patches | L0 / SLC / GRD patches | Label source |
|
| 24 |
+
|---|---|---|---|---|
|
| 25 |
+
| [vessel-detection-dataset](https://huggingface.co/datasets/opensar-insight/vessel-detection-dataset) | 15 | 4,047 | 8,094 each | SARFish / xView3 |
|
| 26 |
+
| [flood-detection-dataset](https://huggingface.co/datasets/opensar-insight/flood-detection-dataset) | 10 | 4,104 | 8,208 each | Kuro Siwo |
|
| 27 |
+
| [rfi-detection-dataset](https://huggingface.co/datasets/opensar-insight/rfi-detection-dataset) | 21 | 1,047 | 2,094 each | Aresys |
|
| 28 |
|
| 29 |
+
L0, SLC and GRD counts are double the unique patch count because each patch is provided in both polarisations. Flood and RFI datasets include segmentation masks.
|
| 30 |
|
| 31 |
+
Vessel and flood patches are 512 × 512 px. RFI patches range from roughly 1400 × 1400 to 1600 × 1600 px, as they follow SLC burst boundaries. Train, validation and test splits are provided for every use case.
|
| 32 |
|
| 33 |
+
Coverage spans the Gulf of Guinea, Iceland, Denmark, Norway and Italy for vessel detection; Albania, Madagascar, India, Nicaragua, Pakistan, the Philippines, the United States, Lithuania, Sri Lanka and Peru for flood detection; and sites across Europe, the Middle East, East Asia, Africa and the Americas for RFI.
|
| 34 |
|
| 35 |
+
## Models
|
|
|
|
| 36 |
|
| 37 |
+
For the vessel detection and RFI detection use cases, we developed two families of ML models: a large, unrestricted model for ground-based use, and a small, optimised model designed for onboard deployment. The small vessel detection model is knowledge-distilled from the large one.
|
| 38 |
|
| 39 |
+
- [vessel-detection-model](https://huggingface.co/opensar-insight/vessel-detection-model) — YOLO-based
|
| 40 |
+
- [rfi-detection-model](https://huggingface.co/opensar-insight/rfi-detection-model) — UNet-based
|
| 41 |
+
|
| 42 |
+
## Codebase
|
| 43 |
+
|
| 44 |
+
The full codebase is at https://github.com/ESA-PhiLab/OpenSARInsight
|
| 45 |
+
|
| 46 |
+
It covers dataset generation from Sentinel-1 products, SAR preprocessing, model training and validation, and inference, with a Docker environment for reproducibility.
|
| 47 |
+
|
| 48 |
+
The code is released under the MIT License, except for components built on Ultralytics YOLO, which are distributed under AGPL-3.0. Licences for the datasets are given in their individual repositories.
|
| 49 |
+
|
| 50 |
+
## Consortium
|
| 51 |
+
|
| 52 |
+
Indra Space leads the project, supporting the generation of open-source datasets and leading the development of ML models for extracting insights from raw SAR data. The consortium also includes the University of Alcalá (AES3 research group), INTA — the National Institute of Aerospace Technology (Spaceborne SAR Systems and Calibration Group), and ESA's Φ-lab.
|
| 53 |
+
|
| 54 |
+
- [Indra Space](https://space.indragroup.com/en)
|
| 55 |
+
- [University of Alcalá](https://uah.es/es/)
|
| 56 |
+
- [INTA](https://www.inta.es/INTA/es/index.html)
|
| 57 |
+
- [ESA Φ-lab](https://philab.esa.int)
|
| 58 |
+
|
| 59 |
+
## Resources
|
| 60 |
+
|
| 61 |
+
- [Project website](https://opensarinsightweb.web.uah.es/)
|
| 62 |
+
- [Publications](https://opensarinsightweb.web.uah.es/publication.html)
|