Until now, OCR Engine2 always returned the full overlay, even if isOverlayRequired=False was added to the API call. With this update, Engine2 gets support for the isOverlayRequired=False parameter. So then it works exactly as OCR Engine1.
Plus some other smaller improvements and bug fixes.
Initially, this update will be deployed to the free OCR API endpoint. If everything functions as expected, we will proceed to roll out this update to the PRO OCR API endpoints approximately two weeks later.
@Maxim I confirmed the issue. Workaound: You can solve this by adding
isoverlayrequired=true
to your API call. The “isTable=true” table parsing logic needs the overlay data to do its work.
With Engine1 we activate isoverlayrequired=true automatically when isTable=true is used, and with OCREngine2 this automatic activation is missing. We will add it soon. For now, manually adding isoverlayrequired=true is a good workaround.
Update 2024/2/9: This has been fixed. So Engine2 behaves now exactly as Engine1: We internally set isoverlayrequired=true when isTable=true is used.
This is a valid combination (= no OCR API error), but isTable = true overrides isoverlayrequired=false and sets it internally to isoverlayrequired=true. Reason: The overlay data is needed for the table parsing logic.
This was always the case since we introduced the table ocr feature. Nothing has changed in this respect.
What the above updated added was the support for isoverlayrequired=false flag for Engine2.