Azure AI Document Intelligence
An Azure service that turns documents into usable data. Previously known as Azure Form Recognizer.
1,535 questions
Custom Model - Form Recognizer
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(313.6, 33%, 40%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3ESY%3C/text%3E%3C/svg%3E)
Saleem Yusuff
1
Reputation point
I have trained a custom model for a specific form, and its able to extract data based on the labels that i have created, they work absolutely fine as long as only 1 page is uploaded, when i upload the same form multiple pages then its reading data only out of the first page, how do i extract data from the other pages as well.
Ex: A form containing Customer Information - trained using Custom Model.
If i upload a pdf that has only 1 page its extracting the data correctly, when i upload a pdf that has 10 pages of the same form type with different customer information its reading only the first page and producing the output but not of the remaining pages.
What am i doing wrong here??
{count} votes
-
romungi-MSFT 43,696 Reputation points Microsoft Employee2022-04-05T15:35:37.617+00:00 @Saleem Yusuff Are you on a free tier for your form recognizer resource? If yes, then the call to analyze with your custom model id will only process the first two pages irrespective of the number of pages in the document.
If you are not using free tier and upgraded to standard tier then the JSON response should contain pages section which will contain the text extracted from all the pages.
-
Saleem Yusuff 1 Reputation point
2022-04-14T07:37:27.027+00:00 No , the Json is also not returning the values, only the first page data is getting extracted, and yes i am not on free tier so the 2 page rule is not an issue, i am trying to understand how to do batch processing using form recognizer, if i upload a pdf file that contains multiple pages and want each page to be extracted individually so that i can insert the data from each page in to a database
-
romungi-MSFT 43,696 Reputation points Microsoft Employee
2022-04-18T13:06:41.13+00:00 Ok, the result response should be of the format mentioned in this reference. pages and pageNumber should provide analyze result for all the pages of the form.
Based on the reported behavior it seems like OCR might be failing with the rest of the pages. Have you tried to train any other model to check if same scenario is seen?{ "status": "succeeded", "createdDateTime": "2022-01-30T12:42:07Z", "lastUpdatedDateTime": "2022-01-30T12:42:13Z", "analyzeResult": { // Basic analyze result metadata "apiVersion": "2022-01-30-preview", // REST API version used "modelId": "prebuilt-invoice", // ModelId used "stringIndexType": "textElements", // Character unit used for string offsets and lengths: textElements, unicodeCodePoint, utf16CodeUnit "content": "CONTOSO LTD.\nINVOICE\nContoso Headquarters...", "pages": [ // List of pages analyzed { // Basic page metadata "pageNumber": 1, } ] } }
Sign in to comment