Umbra Space is a 9-year-old company that manufactures and operates a Synthetic Aperture Radar (SAR) satellite fleet. These satellites can capture high-resolution images day or night, even in challenging weather conditions. Umbra recently released almost 1,000 SAR satellite images displaying ships from around the world. They also have an open data program and have made these images available in various formats. The author of the text trained a ship detection model using Ultralytics’ YOLOv5 and the High-Resolution SAR Images Dataset (HRSID). They provide details about their workstation, software installation, and the process of training the model. They also discuss the inference results and share visualizations of the detected ships in different locations.
| Signal | Change | 10y horizon | Driving force |
|---|---|---|---|
| Umbra releases satellite images of ships | Availability of ship detection data | Increased availability of ship detection imagery | Increased demand for ship tracking and monitoring |
| Training ship detection model with SAR imagery | Adoption of SAR imagery for ship detection | Improved accuracy in ship detection models | Improved technology and algorithms for ship detection |
| High-resolution SAR imagery dataset (HRSID) | Availability of high-resolution SAR imagery dataset | Increased availability and quality of SAR imagery dataset | Advancements in SAR satellite technology |
| Training ship detection model with YOLOv5 | Adoption of YOLOv5 for ship detection | Improved accuracy in ship detection models | Improvements in deep learning algorithms and frameworks |
| Inferring ship detections in Umbra’s imagery | Ship detection in real-time satellite images | Real-time ship detection in satellite imagery | Increased efficiency in ship detection algorithms |
| Visualizing ship detections in satellite imagery | Visualization of ship detections in satellite images | Improved understanding of ship patterns and distributions | Enhanced tools and software for visualization |