All products
Object Detection
Find and classify real-world assets in imagery, as map features.
object-detection
The problem
Asset inventories go stale the moment they are finished. Field survey is slow and expensive, and the imagery that would answer the question — satellite, aerial, street-level — is usually already sitting on a server, unread.
How it works
Computer vision models detect and classify objects in imagery, then place each detection in real-world coordinates. Detections carry a class, a confidence, and a link back to the source frame, so anything uncertain can be reviewed instead of quietly trusted.
Inputs
- Street-level and panoramic imagery
- Aerial and satellite imagery
- Drone captures
- Existing asset inventory, if you have one
Outputs
- Georeferenced detections as point or polygon features
- Class, confidence, and source frame per detection
- Inventory tables and coverage statistics
- A review queue for low-confidence cases
Applications
- Road and traffic sign inventory
- Street furniture and urban asset mapping
- Utility and network asset detection
- Building and rooftop extraction
- Road condition and marking surveys
In your GIS
Detections arrive as GIS features that can be compared against your existing inventory to produce a difference list: missing, moved, added.