Azure Document Intelligence Python SDK - Barcodes feature not working

Question

Azure Document Intelligence Python SDK - Barcodes feature not working

Yi Qiang 0

We are calling the document intelligence API using the python SDK and a custom model. Everything seems to work fine, except that the barcodes feature is not working. However, analyzing the same photo using the same model Azure Document Intelligence Studio produces the expected result.

Sample code:

  poller = await document_intelligence_client.begin_analyze_document( 
    model_id=model_id,
    analyze_request=analyze_request,
    features=[DocumentAnalysisFeature.OCR_HIGH_RESOLUTION, DocumentAnalysisFeature.BARCODES],
  )

When we iterate for the pages array of the result and check for any barcodes, it reports None.

Name: azure-ai-documentintelligence

Version: 1.0.0b4

Rohit Mungi 49,116 Reputation points Microsoft Employee Moderator

2024-10-23T05:27:39.5333333+00:00
@Yi Qiang I think with the studio the custom model operation is actually using the REST API with the parameter DocumentAnalysisFeature.BARCODES as part of analyze operation. Since you are using the SDK along with additional feature, I am not sure if that could cause an issue. As per the sample from SDK repo only prebuilt-layout is documented for additional feature, could you check first with your custom model id with only BARCODES feature and then use the prebuild-layout model as model id if custom model does not provide any result?

with open(path_to_sample_documents, "rb") as f: poller = document_intelligence_client.begin_analyze_document( "prebuilt-layout", analyze_request=f, features=[DocumentAnalysisFeature.BARCODES], content_type="application/octet-stream", ) result: AnalyzeResult = poller.result()

Rohit Mungi 49,116 Microsoft Employee Moderator

@Yi Qiang I ran a test with an image as URL source and it worked to retrieve the barcodes. Here is my analyze request.

    formUrl = "URL of image I referenced above"

    custom_model_id = "barcode"



    poller = document_intelligence_client.begin_analyze_document(
        model_id=custom_model_id, 
        analyze_request=AnalyzeDocumentRequest(url_source=formUrl), features=[DocumentAnalysisFeature.OCR_HIGH_RESOLUTION, DocumentAnalysisFeature.BARCODES]     )

    
    result = poller.result()


   # Iterate over extracted barcodes on each page.
    for page in result.pages:
        print(f"----Barcodes detected from page #{page.page_number}----")
        if page.barcodes:
            print(f"Detected {len(page.barcodes)} barcodes:")
            for barcode_idx, barcode in enumerate(page.barcodes):
                print(f"- Barcode #{barcode_idx}: {barcode.value}")
                print(f"  Kind: {barcode.kind}")
                print(f"  Confidence: {barcode.confidence}")
                if not barcode.polygon:
                    print("  Bounding regions: N/A")
                else:
                    print("  Bounding regions: ")
                    print(
                        ", ".join(
                            [
                                f"[{barcode.polygon[i]}, {barcode.polygon[i + 1]}]"
                                for i in range(0, len(barcode.polygon), 2)
                            ]
                        )
                    )

    print("----------------------------------------")

The response, detects 2 barcodes.

----Barcodes detected from page #1----
Detected 2 barcodes:
- Barcode #0: http://contoso.com/orders/100
  Kind: DocumentBarcodeKind.Q_R_CODE
  Confidence: 1
  Bounding regions:
[1192, 385], [1529, 388], [1526, 722], [1189, 719]
- Barcode #1: 012345678905
  Kind: DocumentBarcodeKind.UPCA
  Confidence: 1
  Bounding regions:
[472, 1134], [765, 1134], [765, 1322], [472, 1322]
----------------------------------------

I am also using version 1.0.0b4

Yi Qiang 0 Reputation points

2024-10-23T15:30:12.6766667+00:00

Hi, thanks for the quick response. I was able to get it working by using the prebuilt-layout model. Is it not possible to extract barcodes using a custom model, then?
Rohit Mungi 49,116 Reputation points Microsoft Employee Moderator

2024-10-24T06:04:17.9966667+00:00

@Yi Qiang I used a custom model in my previous response, and it worked with the script provided. I have pretty much used the same request parameters but used URL instead of local image. Could you try that approach too?
Rohit Mungi 49,116 Reputation points Microsoft Employee Moderator

2024-10-25T10:35:36.46+00:00

@Yi Qiang Did you get a chance to try with a custom model with the sample i provided? Thanks!!

Your answer

Rohit Mungi 49,116 Reputation points Microsoft Employee Moderator

2024-10-23T05:27:39.5333333+00:00

@Yi Qiang I think with the studio the custom model operation is actually using the REST API with the parameter DocumentAnalysisFeature.BARCODES as part of analyze operation. Since you are using the SDK along with additional feature, I am not sure if that could cause an issue. As per the sample from SDK repo only prebuilt-layout is documented for additional feature, could you check first with your custom model id with only BARCODES feature and then use the prebuild-layout model as model id if custom model does not provide any result?

with open(path_to_sample_documents, "rb") as f: poller = document_intelligence_client.begin_analyze_document( "prebuilt-layout", analyze_request=f, features=[DocumentAnalysisFeature.BARCODES], content_type="application/octet-stream", ) result: AnalyzeResult = poller.result()
Rohit Mungi 49,116 Reputation points Microsoft Employee Moderator

2024-10-23T10:36:48.22+00:00

@Yi Qiang I ran a test with an image as URL source and it worked to retrieve the barcodes. Here is my analyze request.

formUrl = "URL of image I referenced above" custom_model_id = "barcode" poller = document_intelligence_client.begin_analyze_document( model_id=custom_model_id, analyze_request=AnalyzeDocumentRequest(url_source=formUrl), features=[DocumentAnalysisFeature.OCR_HIGH_RESOLUTION, DocumentAnalysisFeature.BARCODES] ) result = poller.result() # Iterate over extracted barcodes on each page. for page in result.pages: print(f"----Barcodes detected from page #{page.page_number}----") if page.barcodes: print(f"Detected {len(page.barcodes)} barcodes:") for barcode_idx, barcode in enumerate(page.barcodes): print(f"- Barcode #{barcode_idx}: {barcode.value}") print(f" Kind: {barcode.kind}") print(f" Confidence: {barcode.confidence}") if not barcode.polygon: print(" Bounding regions: N/A") else: print(" Bounding regions: ") print( ", ".join( [ f"[{barcode.polygon[i]}, {barcode.polygon[i + 1]}]" for i in range(0, len(barcode.polygon), 2) ] ) ) print("----------------------------------------")

The response, detects 2 barcodes.

----Barcodes detected from page #1---- Detected 2 barcodes: - Barcode #0: http://contoso.com/orders/100 Kind: DocumentBarcodeKind.Q_R_CODE Confidence: 1 Bounding regions: [1192, 385], [1529, 388], [1526, 722], [1189, 719] - Barcode #1: 012345678905 Kind: DocumentBarcodeKind.UPCA Confidence: 1 Bounding regions: [472, 1134], [765, 1134], [765, 1322], [472, 1322] ----------------------------------------

I am also using version 1.0.0b4
Yi Qiang 0 Reputation points

2024-10-23T15:30:12.6766667+00:00

Hi, thanks for the quick response. I was able to get it working by using the prebuilt-layout model. Is it not possible to extract barcodes using a custom model, then?
Rohit Mungi 49,116 Reputation points Microsoft Employee Moderator

2024-10-24T06:04:17.9966667+00:00

@Yi Qiang I used a custom model in my previous response, and it worked with the script provided. I have pretty much used the same request parameters but used URL instead of local image. Could you try that approach too?
Rohit Mungi 49,116 Reputation points Microsoft Employee Moderator

2024-10-25T10:35:36.46+00:00

@Yi Qiang Did you get a chance to try with a custom model with the sample i provided? Thanks!!

Share via

Azure Document Intelligence Python SDK - Barcodes feature not working

Your answer