I am Indian citizen.In our country peoples having different naming formats.According to religion,peoples having names which are differ in length.In some religion,first name consists of only one word while in other religion first name consists of two words.
I have a template in FlexiLayout for credit card application form in which Name field is in horizontal direction.In Name field I want to define First Name,Middle Name,Last Name.When I export this into FlexiCapture then this template matching values for only single document.If I add another document to same document definition then I get junk values.If I add the same image in FlexiLayout and try to map the First Name value then the mapping of previous document is disturbed because First Name length differ on both the documents.
Please refer the below screenshot.
I have a template in FlexiLayout for credit card application form in which Name field is in horizontal direction.In Name field I want to define First Name,Middle Name,Last Name.When I export this into FlexiCapture then this template matching values for only single document.If I add another document to same document definition then I get junk values.If I add the same image in FlexiLayout and try to map the First Name value then the mapping of previous document is disturbed because First Name length differ on both the documents.
Please refer the below screenshot.
Comments
15 comments
Reason for such behavior is probably somwhere in flexilayout structure or recognition settings.
Is it possible to provide flexilayout project and sample images for further investigation?
Regards
Tim
Thanks for responding to my message.I appreciate!
Can you please tell me,where I can send project and images?
I have challenge to upload project through this link.However my project size is almost half of the maximum limit of 10 MB.
Please find error screenshots while uploading.
Please reply me.
Regards
Tim
I have below FTP link.Whether I can put on that.Is it OK?
ftp://ftp.abbyy.com/
please upload your project and images to this location:
ftp://ftp.abbyy.com/f/O8YP6JW7WGT4XRMJYYCTOG05
Sorry for the delay.Because yesterday I was about to go home when I got your message from office at 18:00 IST.Therefore today I am sending project.
Please find on FTP.:cool:
I have put project "abbyy Project" on FTP.Please find.
I have checked provided flexilayout here are some tips to increase recognition quality:
1.Avoid using relations dialogue in field properties. Reason why this approach should be avoided is because when you specify an offset in dots or pixels for regions' borders they will be matched correctly only if all images are of the same size and resolution. A more robust approach is to use advanced propertioes and flexilayout language to specify border regions relatively to found elements like so:
above:somelement.rect.top+someotherelemnt.rect.wod th.
By using such approach you won't be bound to a specific image resolution.
2.Use character string elements or ImageCollection to capture handwritten names
Since first names can consist of two words and may have diffrent length it will not be possible to identify regions for them correctly without use of additional elements.
Possible approach is to use imagecollection objects. If last name consists of only one word you can use this to look for the rightmost image collection, then proceding to middle name and lastly capturing what is left which should be the first name.You can then use these image collections for your output region, or use additional elements to further adjust output regions.
Regards
Tim
I do not understand at all.Can you explain in brief please.
Ive uploaded a sample project to the ftp folder mentioned earlier, file name is OCR_APPFORM Sample.zip
For demonstration purposes i have only left three elements:
Name
FirstName
MiddleName
LastName
Header logic is fine, chosen text is reliable and the only correct option to use as header.
Static Text "Name", everything also seems to be fine.
Region elements "FirstName" "LastName" "MiddleName", by opening their properties and going to relations tab we see that offset parameters are specified in dots. Image resolution is 600 DPI, if you try to process images with diffrent resoluition the offset values will still be as specified but since image size is actually diffrent you will get incorrect results.In the project i have added a page with DPI 300, as you can see mathcing results are absolutely diffrent.
So, what we will do is disable all relations for now.
In order to find last name, we need an additional element of characterstring type-csLastName
In advanced pre-search relations you can see how search area can be spsecified by using flexilayout language, there are no constants so it is irrelevant whether pages are of 600 or 300 dpi.
Additional character strig to find middle name, and limit its right border by previously found last name.
Next wee need a left border for first name, stDr.
Finally add character string to capture anything thats left.
By testing this layout structure on an image with diffrent DPI or size you will see that results are the same due to the fact that we didnt bind to any specific values.
Regards
Timur
I appreciate your detailed response.We have now understand the logic but not up to the mark.Because on the basis of Advanced pre-search relations that you have given is not sufficient for understanding the concept because orientation of Name field is in horizontal direction.So there is no complex logic for Name field.
Certain fields on document such as DOB,PAN No.,Mobile No.,Address requires some complex logic.
So I request you to please provide Advanced pre-search relations for fields DOB,PAN No.,Mobile No.,Address
We have tried ourselves for DOB field but unable to fix.For your understanding,please find screenshot of what we have tried.
Any update?
Please sign in to leave a comment.