Skip to content
Deep Learning - Boat Detection
In this project, I created a boat detection feature layer from high-resolution aerial imagery of Tuborg Havn in Copenhagen using the Text SAM GeoAI model in ArcGIS Pro. The model used a text prompt, “boat,” to automatically identify boats in the imagery and generate polygon features showing their approximate outlines. After the initial detection, I reviewed the output and created a cleaner version of the layer by filtering out weaker detections and very small features that were likely false positives. This made the final result more reliable for interpretation and analysis
Figure 1. This shows the boat detection workflow in ArcGIS Pro, including the Tuborg Havn imagery, the detected boat polygons, and the attribute table used to review detection results. The red outlined polygons show the initial boat detections generated with the Text SAM model, while I used the attribute table to inspect values for size and shape before refining the results.
Figure 2. This figure shows the final cleaned boat detection layer after the initial results were refined to remove weaker and less reliable detections. The remaining red outlined polygons represent the boats kept in the final dataset, creating a more accurate layer for counting boats.
The final boat detection result was a cleaner version of the original output after the weaker detections and small false positives were removed. This left a more accurate set of detected boats in the marina area. Overall, the project showed how Text SAM can be used in ArcGIS Pro to detect boats from aerial imagery and then refine the results into a clean final layer.
Want to print your doc?
This is not the way.
This is not the way.
Try clicking the ··· in the right corner or using a keyboard shortcut (
CtrlP
) instead.