Ship Detection (RGB)

Deep learning model to detect ships in high-resolution satellite imagery.

Pretrained Model Outputs

Geography - US

Lake Worth

Esri, DigitalGlobe, Earthstar Geograpics, CNES/Airbus DS, USDA FSA, USGS, Aerogrid, IGN, IGP, and the GIS User Community
Powered by Esri
Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri

Branford

Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri
Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri

Lighthouse Point

Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri
Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri

Los Angeles

Hammond

Emeryville

San Francisco

Lake Worth

Boca Raton

Aventura

Miami

Lake Worth

Cape Canaveral

Galveston

Geography - Global

Dubai

Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri
Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri

South Africa

Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri
Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri

Russia

Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri
Esri, Maxar, Earthstar Geographics, and the GIS User Community
Powered by Esri

Russia

India

India

Argentina

Australia

New Zealand

Brazil

Model Overview

Input High-resolution, 3-band RGB satellite imagery with a spatial resolution of 30 centimeters, or 0.3 meters.

Output Feature layer with polygons representing the detected ship(s) in the input imagery

Applicable geographies The model is expected to work well in the United States and similar geographies.

Model architecture This model uses the  MaskRCNN  model architecture implemented in the ArcGIS API for Python.

Accuracy metrics This model has an average precision score of 0.683.

Training data The model has been trained on an in-house ship detection dataset.

Limitations 1. The model will work on ship lengths in the range of 12–80 m. Ships less than 12 meters long may or may not be detected. 2. Traditional boats and cargo ships may or may not be detected.  3. The model may detect false positives on land surfaces. Apply a water mask to eliminate such detections.

***********************************