[Python] How to reconstruct an Arabic word from YOLO predictions using class IDs and x-center...

I’m using a YOLO model to detect Arabic characters in an image.

The model gives me two lists after inference:

• classes: a list of predicted class IDs (each ID corresponds to an Arabic character, e.g. 0 = 'ا', 1 = 'ب', etc.)
• centers_x: the x-center coordinates of each detected character

I want to reconstruct the full Arabic word from these predictions from right to left (because Arabic is RTL).
That means: I should sort the characters by their x_center from highest to lowest, and then map the class ID to the corresponding Arabic letter.

Example:

id_to_char = {0: 'ا', 1: 'ب', 2: 'ت'}
classes = [2, 0, 1]
centers_x = [300, 250, 100]


Desired result: 'بتا'

Continue reading...