Deep learning model to detect ships in high-resolution satellite imagery. A brief summary of the item is not available. Add a brief summary about the item.
Deep learning package
by
Item created: Nov 15, 2023 Item updated: Feb 4, 2025 Number of downloads: 2,064
Description
Ship detection plays an important role in defense and security, port management, environmental monitoring, insurance and risk assessment, and maritime search and rescue operations. While Automated Identification System (AIS) is commonly used for ship detection, it has limitations, such as incomplete ship traffic information due to AIS being switched off or malfunctioning. Also, not all ships are equipped with AIS transponders. Satellite imagery-based detection overcomes these limitations.
This pretrained model detects and localizes ships in high-resolution optical satellite images. It can handle both dense and sparse ship patterns. The detected polygons align with the ships' rotation angles to provide more accurate and precise information about the ship's orientation and enhance the spatial representation of ships. The model is based on the MaskRCNN architecture implemented using ArcGIS API for Python and has been trained on an in-house ship detection dataset that covers various ship types and sizes across the United States.
Using the model
Follow the guide to use the model. Before using this model, ensure that the supported deep learning libraries are installed. For more details, check Deep Learning Libraries Installer for ArcGIS.
Fine-tuning the model
This model can be fine-tuned using the Train Deep Learning Model tool. Follow the guide to fine-tune this model.
Input
High-resolution, 3-band RGB satellite imagery with a spatial resolution of 30 centimeters (0.3 meters). Consider adjusting the cell size to approximately 15 centimeters for detecting smaller boats and increasing it for larger boats.
Output
Feature class with polygons representing the detected ship(s) in the input imagery.
Applicable geographies
The model is expected to work well in the United States, Indian Subcontinent and similar geographies.
Model architecture
This model uses the MaskRCNN model architecture implemented in ArcGIS API for Python.
Accuracy metrics
This model has an average precision score of 0.6953.
Training data
This model has been trained on an Esri proprietary ship detection dataset.
Limitations
1. The model will work on ship lengths in the range 12 – 80m. Ships less than 12m 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 over land surfaces. Apply a water mask to eliminate such detections.
Sample results
Here are a few results from the model. To view more, see this story.
An in-depth description of the item is not available.
Layers
Tools
Tables
Basemap
Project Contents:
Solution Contents
Contents
Layers
Screenshots
Terms of Use
No special restrictions or limitations on using the item's content have been provided.
Details
Dashboard views: Desktop
Source:
Creating data in:
Published as:
Other Views:
Dependent items in the recycle bin
Applicable: 2d
Size: 157.234 MB
ID: 2fd653cd9de446ccbab34f69e9e70d81
Image Count: 0
Image Properties
Layer Drawing
Using tiles from a cache
Dynamically from data
Share
Owner
Folder
Categories
This item has not been categorized.
Credits (Attribution)
No acknowledgements.Esri
Comments (7)
Hi. Maybe someone could help me. To work with deeo learning I will always need a pretrained model of the oibjet I want to see, for example If I need to count crocodiles in a river, Do I need a pretrained model to detect crocodiles aparte de la ortofoto? Thanks
Dear Karen, Yes, to count crocodiles in a river using deep learning, you will need a crocodile pretrained model to run on your raster dataset. You can either train your own model in ArcGIS Pro using your dataset to generate a pretrained model for your specific use case and geography, or you can use the already available pretrained models on ESRI Living Atlas. Please follow this link to learn to train a model: https://pro.arcgis.com/en/pro-app/latest/tool-reference/image-analyst/train-deep-learning-model.htm and please feel free to let us know if you face any other issues or have any other query. Thanks , Neha
Thank you Neha and Pathak for your prompt responses. In all pretrained models the way I understood was sample images were provided as I do not have relevant images for each applications. Anyway now I understand .that .doll files do not come with imageries. Thank you so much for clarifying it. Best Sumith
You are welcome. Please feel free to let us know if you face any other issues or have any other query. Thanks Neha
Hello, How do i add the raster image 3 bands into ArcGIS pro content page. I cannot get that from DLPK file. Thanks. Sumith
Dear Sumith, you can add your raster using 'Add Data' tool in the 'Map' tab in ArcGIS Pro by browsing to the location where the raster is saved in your system or you can just drag and drop the raster in the 'Contents' panel. Please feel free to let us know if you face any other issues or have any other query. Thanks Neha
Hi Sumith, DLPK file stands for Deep Learning Package which is a pretrained model to detect ships, but it doesn't include the raster image. You will need to provide your own high resolution 3-band RGB raster (30 cm spatial resolution) for the model to process.