Form Recognizer
-
How to label a table in a form?
In the Azure form recognizer official website "https://azure.microsoft.com/en-in/services/cognitive-services/form-recognizer/" few examples have tables in the sample file. In output also we have an attribute called table in sample json. Please guide us in labeling a table in a custom layout form.
15 votesForm Recognizer discovers and extracts tables automatically. Table results are part of the pageResults section in the JSON output. If the table in the form was not discovered you can label tables a values by labeling each table cell and training with the maximum number of rows in the tables. Form Recognizer does not yet support labeling tables as tables.
-
How to label an electronic generated form for custom training?
After going through steps mentioned in the below source url "https://docs.microsoft.com/en-us/azure/cognitive-services/form-recognizer/quickstarts/label-tool" we didn't get all the products tagged as expected(Refer Output.jpg).
In the output file the highlighted fields were not detected can anyone please suggest us the way to tag the data and to improve the model accuracy.
We have trained the model with 10 samples of same template in different variations.
For Example one form having 2 products in the Description table and another having 6 products and so on.
1 voteCan you please try and label 5 forms with the maximum number of rows you would like to extract ? For example all labeled data have 6 rows and when analyzing extract any number of rows forms.
-
Prebuilt receipts doesn't recognize logo'd merchant names
If a receipt only has the merchant logo and not the merchant name on a receipt, the prebuilt receipt model returns nothing for the MerchantName.
For example, Home Depot has their logo and "The Home Depot" in reverse font on their logo but the OCR doesn't pick it up.2 votes -
Prebuilt receipts doesn't recognize invoice style receipts
Many businesses have 8.5x11 invoice style receipts (parts stores, rental businesses, service businesses, etc.). The prebuilt receipts model returns no data on these even though they are accepted with no errors.
2 votesForm Recognizer pre-built receipts model supports US-Sales receipts such as stores, restaurants etc. To extract data from invoices please train a custom model.
-
Support well known tax forms as pre-built form
Start supporting well known tax forms i.e. w2, 1099 forms etc as pre-built forms so users don't have to create custom model for it and train them.
4 votes -
Add the ability to train a model with a blob larger than 4mb
I am trying to use Microsoft Form Recogniser to get the key value pairs from medical forms. However there are a number of different types of form and cannot currently train an accurate enough model to use with the current size limit.
10 votesForm Recognizer v2.0 (preview) enables training large data-sets and analyzing large documents.
-
Define a key/value when its wrong.
I make several test and works very fine but in some cases the key value is wrong even the key is on de left of the value, it would be nice to specify the key when its wrong to train a better model.
1 voteForm Recognizer v2.0 (preview) provides the option to label forms and train a model to extract the values of interest.
-
Handwritten forms
We need an ability to extract key-value pairs from a form which is hand filled or hand written. For example - an insurance claim form which is filled in by a person or an agent. Is some work being done on this ?
3 votes -
Form with tables, checkboxes, and whole lot of square boxes
I have trained my form recognizer model using all the requirements and tips provided in documentation. However, the output is not good at all.
The form that I am trying to train on is attached (empty version). I have used 2 filled in forms as well as per requirements.
The form has two small tables at the top - even these haven't been read correctly. And then there are whole lot of square boxes (representing one character) to fill in details of the users. The boundaries of these square boxes are being read either as 1 or as a dash…8 votes -
Form Recognizer doesn't work with Japanese characters even with very basic files.
The attached example is used for training, but the trained model is not able to segregate keys and values. All the keys are set-up to "Tokens" value ( eg: "0": [ "Tokens" ])
Analyze Forms component will only extract some of the data, and will anyway allocate all of them on only one key.
Is there a version of Form Recognizer that works better with Japanese characters ? I am currently using the one on West Europe.
1 vote -
Does Form Recognizer work with files in Blob Storage
Having an issue with the QuickStart for Form Recognizer .NET SDK.
The example works great when I feed it a PDF file that is local to my PC. I want to pass it a PDF file that exists in Azure Blob Storage so I modified the example to use a MemoryStream variable instead of a FileStream. This give me a "415 - UnsupportedMediaType" error message even though the same file works correctly if I analyze it from my local PC.
What am I missing? Does Form Recognizer support analyzing files from blob storage?
1 voteForm Recognizer v2.0 (preview) supports analyzing files from blob storage.
-
Better detection of field values wihtin character boxes
Our forms (or our client forms, which we have no control over) use the style of value field where there are a series of boxes, and every character must be entered into a box. I believe these are designed for template-based form recognition, but this seems to be a big problem for Form Recognizer.
2 votes -
Recognize Checkbox and Signature sections
We have a number of paper forms that are completed by staff. The forms include checkboxes. We need to confirm that the checkboxes have been checked.
Additionally, we require the form to be signed by the person completing it. We'd like to confirm that a mark has been made in the signature area.
It would be great if there was a way that Form Recognizer could just register that a pen mark has been made in a certain area, even if it can't read it (e.g. a signature or a tick in a box)
31 votesThank you for the request. This checkbox feature is being planned for the next release.
-
Increase Timeout on Training using PDF and Logic Apps
When I use logic apps to train a model, it seems to timeout when I use more than 5 PDFs (I used 10). 5 PDFs seemed to work at 52 seconds.
2 votesForm Recognizer Logic Connector currently uses the Form Recognizer v1.0 API which is a sync API and has a HTTP timeout limit. Form Recognizer v2.0 (preview) is an Async API and does not have this timeout limit. You can create a custom Logic App connector to Form Recognizer v2.0 (preview) API.
-
Tips for handling key value pairs from Form Recognizer Json output
Can you offer any practical pointers, thoughts or guidance on ways to deserialize the Json output please? The structure of the key/value pairs is understandable and has helpful data like BoundingBoxes but it is complex to deserialize?
2 votes -
Fixed: Microsoft.Azure.CognitiveServices.FormRecognizer.Models.ErrorResponseException: ‘Operation returned an invalid status
TrainCustomModelAsync: Microsoft.Azure.CognitiveServices.FormRecognizer.Models.ErrorResponseException: ‘Operation returned an invalid status code ‘RequestEntityTooLarge’
We have followed code examples but we are getting the following error (FR1.png) when calling TrainCustomModelAsync: Microsoft.Azure.CognitiveServices.FormRecognizer.Models.ErrorResponseException: ‘Operation returned an invalid status code ‘RequestEntityTooLarge’
We are trying to pass a reference to the Training Data that has been populated in Azure (FR2.png). It is a collection of about 5 pdfs. Each with approx 4 pages at 300dpi or so.
We adapted from the following sample: https://docs.microsoft.com/en-us/azure/cognitive-services/form-recognizer/quickstarts/dotnet-sdk
Any help is gratefully received.
3 votesForm Recognizer v2.0 (preview) provides an async API to train large datasets and analyze large documents.
-
Add support for Cyrillic
I would like to see support for Cyrillic-based languages, such as Bulgarian, Russian, etc.
1 vote -
Increase limit of number of samples for custom training
We understand that the number of samples for custom training is currently limited to 50. We hope that by providing more samples, the model would be more accurate in the end on large volumes, so would it be possible to raise that limit?
8 votesThank you for your request, this feature has been implemented: “The total size of the training data set must be 500 pages or less.”
https://docs.microsoft.com/en-us/azure/cognitive-services/form-recognizer/build-training-data-set -
Issue on fields with multiple line headers
First, congrats to the team for the great new product!
It works already quite well on the forms we tried, despite small variation of templates for same type of document, which is a great help!
But at the time being, we can't consider to use it, as it seems to have an issue with fields having headers split on multiple lines. We can provide samples but not on a public forum.5 votes -
Manual Training
Make it trainable (maybe with a UI) where users can manually point or map fields they need. We have tons of different templates of Invoices but we basically need only the main fields on the Invoice recognized, and currently we have no way of pointing out what we need and with the 50 pages\4 Mb per model and we feel very limited.
Maybe a way of saying Invoice #, or Invoice No. or Inv# should be mapped to Invoice Number for all of the results for this model; this would be of much help when extracting data from different Invoice…
29 votesForm Recognizer v2.0 (preview) enables labeling forms and training a model to extract the values of interest. Train a Form Recognizer model with labels using the sample labeling tool.
https://docs.microsoft.com/en-us/azure/cognitive-services/form-recognizer/quickstarts/label-tool
- Don't see your idea?