Merging Detection and Recognition Classes into a unified class?
I wonder if the MMOCR developers have thought about combining the detection and recognition classes into one unified class. If they have, it might be more efficient to call the recognizer to identify text immediately upon the detector detecting a text region within an image. Currently, the recognizer waits until the detector accumulates all text regions, resulting in a distinct workflow.
Merging Detection and Recognition Classes into a unified class?
I wonder if the MMOCR developers have thought about combining the detection and recognition classes into one unified class. If they have, it might be more efficient to call the recognizer to identify text immediately upon the detector detecting a text region within an image. Currently, the recognizer waits until the detector accumulates all text regions, resulting in a distinct workflow.
Any other context?
No response