Inconsistent angle OCR Read API
Anonymous
I am using the Azure Cognitive Services OCR Read API for document images.
In some of the docs I am observing inconsistent detection of the document angle and correspondingly the bounding boxes are rotated. In the example image show below, majority of the text is upright, with a text line which is rotated. My suspicion is that the rotated text line is causing this issue.
Question is - does the OCR read api have a consistent behaviour? or there are chances where it might show no rotation for some responses, while rotation for other responses for the same image ?
Sign in to answer