Published December 12, 2025
| Version v1
Dataset
Open
A national fish pond dataset for Nigeria generated using automated detection and segmentation
Authors/Creators
-
1.
University of Delaware
- 2. National Space Research and Development Agency, Abuja, Nigeria
-
3.
International Institute of Tropical Agriculture
- 4. WASCAL Graduate Research Program in Climate Change Economics (GRP-CCE), UCAD, Senegal
- 5. ActionAid Nigeria
-
6.
University of Ibadan
-
7.
University of Washington
Description
fine_tune_final_dataset_freeze_v02_2.zip: YOLOv7 segmentation model
all_training_clean_state_clusters15_t01_wmore.zip:
Shape file used to train Random Forest model. The polygons labeled as Fish Pond are the manually annotated ponds. Columns are:
| ids | Unique identifier of object. |
| state | State where the object is located |
| geometry | Polygon geometry |
| split | train/test/valid split |
| Fishpond | True label, 1: Fish Pond, 0: No Fish Pond |
| area | area of the prediction in meters |
| length | perimeter of the prediction in meters |
| inv_isoper | isoperimetric ratio: how round is the prediction, 1 for circle |
| ave_r | average of red color for prediction |
| ave_g | average of green color for prediction |
| ave_b | average of blue color for prediction |
| std_r | standard deviation of red color for prediction |
| std_g | standard deviation of green color for prediction |
| std_b | standard deviation of blue color for prediction |
| MSAVI_1km | Modified Soil Adjusted Vegetation Index for 1 km^2 around the polygon in December 2023 |
| NDBI_1km | Normalized Difference Built-up Index for 1 km^2 around the polygon in December 2023 |
| NDVI_1km | Normalized Difference Vegetation Index for 1 km^2 around the polygon in December 2023 |
| NDWI_1km | Normalized Difference Water Index for 1 km^2 around the polygon in December 2023 |
| MSAVI_pond | Modified Soil Adjusted Vegetation Index for the area of the polygon in December 2023 |
| NDBI_pond | Normalized Difference Built-up Index for the area of the polygon in December 2023 |
| NDVI_pond | Normalized Difference Vegetation Index for the area of the polygon in December 2023 |
| NDWI_pond | Normalized Difference Water Index for the area of the polygon in December 2023 |
| MSAVI_01 | Modified Soil Adjusted Vegetation Index for 1 km^2 around the polygon in January 2024 |
| NDBI_01 | Normalized Difference Built-up Index for 1 km^2 around the polygon in January 2024 |
| NDVI_01 | Normalized Difference Vegetation Index for 1 km^2 around the polygon in January 2024 |
| NDWI_01 | Normalized Difference Water Index for 1 km^2 around the polygon in January 2024 |
| MSAVI_stat | Modified Soil Adjusted Vegetation Index for the state in December 2023 |
| NDBI_state | Normalized Difference Built-up Index for the state in December 2023 |
| NDVI_state | Normalized Difference Vegetation Index for the state in December 2023 |
| NDWI_state | Normalized Difference Water Index for the state in December 2023 |
| clus_lab15 | id of cluster, obtained using DBSCAN(eps=15, min_samples=3) |
| clus_siz15 | number of neighbours in cluster, obtained using DBSCAN(eps=15, min_samples=3) |
| pred | Random Forest prediction for Fish Pond or no Fish Pond |
fishpond_rf_model_state_cluster15_ndvi_ts_wmore : Random Forest Model
final_fishponds_states.zip: All predictions made by the segmentation model. One zip file per state. To filter out only the fish ponds, filter by pred=1
Files
all_training_clean_state_clusters15_t01_wmore.zip
Additional details
Software
- Repository URL
- https://github.com/fedhere/FPAI
- Programming language
- Python