Detected Objects to OCR

Overview

Detected Objects to OCR node merges nearby rectangular detections into text-like regions and returns OCR-style contours, boxes, labels, and confidence.

Use this node after a character or text-fragment detector when you want reading-order merge before further processing. For full end-to-end text recognition, see OCR.

Input

Input Image

image required

The image frame associated with the detections. Connect this to a camera or upstream image output.

Detected Objects

array required

List of detected objects to convert into OCR text regions. Each object must include Bounding Box (bbox: [x, y, width, height]) and Label.

Merge Mode

string required

How adjacent objects are merged into text regions.

Values:

• TEXT_HORIZONTAL (default) — horizontal text merge
• TEXT_VERTICAL — vertical text merge
• CUSTOM — custom thresholds and merge direction

When Merge Mode is CUSTOM, the following fields appear:

Merge Threshold

array

Threshold relative to object size. Adjacent objects within the threshold are merged.

• Horizontal (default 1.2, range 0–10)
• Vertical (default 1, range 0–10)

Label Merge Direction

string

Order used to merge labels based on object centers.

Values:

• L_TO_R (default) — left to right
• R_TO_L — right to left
• T_TO_B — top to bottom
• B_TO_T — bottom to top

Output

Overlay Image

image

Frame showing merged OCR-style regions.

Detected Count

integer

Number of merged text regions.

Detected Texts

array

Merged text regions. Each entry contains:

• Points (contour) — region polygon points [x, y]
• Bounding Box (bbox): [x, y, width, height]
• Label — merged label text
• Confidence — weighted average of the merged objects