EASTGeometryType
Enums · AiDotNet.Enums
The shape EAST predicts for each detected piece of text.
For Beginners
When EAST finds text in a picture it has to describe WHERE that text is. There are two ways it can do that, and they suit different pictures:
How It Works
The choice changes the MODEL, not just the reporting: the geometry head predicts five numbers per position for a rotated box (four edge distances plus an angle) and eight for a quadrilateral (four corner offsets). Zhou et al. (CVPR 2017) describe both.
This was previously a string taking “RBOX” or “QUAD”. A misspelling compiled fine and
silently fell through to the five-channel head, producing a model of the wrong shape with no
error – which is exactly why a closed set of values belongs in an enum.
Fields
| Field | Summary |
|---|---|
EASTGeometryType Quad |
An arbitrary quadrilateral: the offsets of its four corners. |
EASTGeometryType RBox |
A rotated rectangle: four distances from the centre to the edges, plus a rotation angle. |
int value__ |