Ship Detection¶

This notebook demonstrates how to use the geoai package for ship detection using a pre-trained model.

Install package¶

To use the geoai-py package, ensure it is installed in your environment. Uncomment the command below if needed.

In [ ]:

# %pip install geoai-py

# %pip install geoai-py

Import libraries¶

In [ ]:

import geoai

import geoai

Download sample data¶

In [ ]:

raster_url = (
    "https://huggingface.co/datasets/giswqs/geospatial/resolve/main/ships_dubai.tif"
)
raster_path = geoai.download_file(raster_url, "ships_dubai.tif")

raster_url = ( "https://huggingface.co/datasets/giswqs/geospatial/resolve/main/ships_dubai.tif" ) raster_path = geoai.download_file(raster_url, "ships_dubai.tif")

Visualize data¶

In [ ]:

geoai.view_raster(raster_url)

geoai.view_raster(raster_url)

Initialize model¶

In [ ]:

detector = geoai.ShipDetector()

detector = geoai.ShipDetector()

Generate masks¶

In [ ]:

output_path = "ships_dubai_masks.tif"

output_path = "ships_dubai_masks.tif"

In [ ]:

masks_path = detector.generate_masks(
    raster_path,
    output_path=output_path,
    confidence_threshold=0.9,
    mask_threshold=0.7,
    overlap=0.25,
    chip_size=(256, 256),
    batch_size=4,
)

masks_path = detector.generate_masks( raster_path, output_path=output_path, confidence_threshold=0.9, mask_threshold=0.7, overlap=0.25, chip_size=(256, 256), batch_size=4, )

Vectorize masks¶

In [ ]:

gdf = detector.vectorize_masks(
    output_path,
    output_path="ships_dubai_masks.geojson",
    confidence_threshold=0.8,
    min_object_area=100,
    max_object_size=10000,
)

gdf = detector.vectorize_masks( output_path, output_path="ships_dubai_masks.geojson", confidence_threshold=0.8, min_object_area=100, max_object_size=10000, )

Visualize initial results¶

In [ ]:

geoai.view_vector_interactive(gdf, column="confidence", tiles=raster_url)

geoai.view_vector_interactive(gdf, column="confidence", tiles=raster_url)

Calculate geometric properties¶

In [ ]:

gdf = geoai.add_geometric_properties(gdf)
gdf.head()

gdf = geoai.add_geometric_properties(gdf) gdf.head()

In [ ]:

geoai.view_vector_interactive(gdf, column="confidence", tiles=raster_url)

geoai.view_vector_interactive(gdf, column="confidence", tiles=raster_url)

Filter results¶

In [ ]:

m = geoai.view_raster(raster_url, backend="ipyleaflet")
m

m = geoai.view_raster(raster_url, backend="ipyleaflet") m

Use the drawing tool to select the area of interest.

In [ ]:

aoi = m.user_rois

if aoi is None:

    aoi = {
        "type": "FeatureCollection",
        "features": [
            {
                "type": "Feature",
                "properties": {},
                "geometry": {
                    "type": "Polygon",
                    "coordinates": [
                        [
                            [55.133729, 25.110277],
                            [55.134072, 25.11393],
                            [55.134823, 25.115601],
                            [55.136025, 25.117116],
                            [55.137677, 25.118127],
                            [55.140145, 25.118787],
                            [55.142248, 25.11902],
                            [55.142012, 25.118243],
                            [55.140831, 25.116728],
                            [55.13948, 25.116903],
                            [55.137956, 25.116825],
                            [55.136132, 25.115543],
                            [55.13566, 25.114416],
                            [55.135467, 25.1136],
                            [55.135939, 25.112609],
                            [55.136218, 25.111657],
                            [55.13551, 25.110685],
                            [55.134373, 25.110102],
                            [55.133729, 25.110277],
                        ]
                    ],
                },
            }
        ],
    }

aoi = m.user_rois if aoi is None: aoi = { "type": "FeatureCollection", "features": [ { "type": "Feature", "properties": {}, "geometry": { "type": "Polygon", "coordinates": [ [ [55.133729, 25.110277], [55.134072, 25.11393], [55.134823, 25.115601], [55.136025, 25.117116], [55.137677, 25.118127], [55.140145, 25.118787], [55.142248, 25.11902], [55.142012, 25.118243], [55.140831, 25.116728], [55.13948, 25.116903], [55.137956, 25.116825], [55.136132, 25.115543], [55.13566, 25.114416], [55.135467, 25.1136], [55.135939, 25.112609], [55.136218, 25.111657], [55.13551, 25.110685], [55.134373, 25.110102], [55.133729, 25.110277], ] ], }, } ], }

In [ ]:

import geopandas as gpd

import geopandas as gpd

In [ ]:

aoi_gdf = gpd.GeoDataFrame.from_features(aoi["features"], crs="EPSG:4326").to_crs(
    gdf.crs
)

aoi_gdf = gpd.GeoDataFrame.from_features(aoi["features"], crs="EPSG:4326").to_crs( gdf.crs )

Intersect the selected area with the vectorized masks to filter the results.

In [ ]:

gdf_filter = gdf[gdf.intersects(aoi_gdf.geometry[0])]
gdf_filter.head()

gdf_filter = gdf[gdf.intersects(aoi_gdf.geometry[0])] gdf_filter.head()

Visualize final results¶

In [ ]:

geoai.view_vector_interactive(gdf_filter, column="confidence", tiles=raster_url)

geoai.view_vector_interactive(gdf_filter, column="confidence", tiles=raster_url)

Save results¶

In [ ]:

gdf_filter.to_file("ships_dubai.geojson")

gdf_filter.to_file("ships_dubai.geojson")