You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Improve the algorithm for deciding what text is present in a region based on three inputs from supported services. Currently, we make many assumptions and use rather straightforward approach. Here are some ideas:
give higher priority if we are expecting a number or if we have a set of expected words/values
work with with individual words and their positions instead of whole lines
if majority says there is one item and one service says its two perhaps the majority is right, but there might be some exceptions
If voting on words fails, by default merge all together, do the pairwise align and vote on letters. Alternatively, choose output from only one service - at least the words will make sense?
The text was updated successfully, but these errors were encountered:
There are cases when text from multiple ROIs is joined into single identified text, with coordinates exceeding ROIs bounds. These should be given a lower priority.
Improve the algorithm for deciding what text is present in a region based on three inputs from supported services. Currently, we make many assumptions and use rather straightforward approach. Here are some ideas:
If voting on words fails, by default merge all together, do the pairwise align and vote on letters. Alternatively, choose output from only one service - at least the words will make sense?
The text was updated successfully, but these errors were encountered: