RedEye Building Annotations

About These Annotations

These annotations represent two independent labeling efforts to identify and segment architectural features in building imagery extracted from 360° panoramic videos. Both datasets use the same source imagery but were annotated by different teams with slightly different approaches.

LYD Subset

Comprehensive architectural annotations labeled by the LYD team, covering 20 classes of building features.

162,315 Total Objects

2,243 Images

48,334 Windows

20 Classes

View LYD Gallery →

Datapure Subset

Detailed architectural annotations labeled by the Datapure team, with extended coverage of decorative elements.

125,496 Total Objects

1,978 Images

41,565 Windows

20+ Classes

View Datapure Gallery →

Annotation Classes

Both datasets cover the following architectural features:

Windows - Various window types and styles
Doors - Entry and exit points
Building facades - Main building structures
Roofs - Roof structures and materials
Balconies - Protruding balcony structures
Sills - Window sills and ledges
Cornices - Decorative crown molding
Columns - Structural and decorative pillars
Molding - Decorative trim work
And more... - Additional architectural details