0

Getting "Can't find any OCR files for training" while training labelled OCR forms using AZURE Form Recognizer. I have the labelled files in my blob.

POST API Request:-

{
    "source":"https://blobstoragefg.blob.core.windows.net/trainsupervised?sp=racwdli&st=2022-12-04T01:40:39Z&se=2022-12-31T09:40:39Z&spr=https&sv=2021-06-08&sr=c&sig=73kKnD88Bx4xOBNKUsdWxbypksIOpYjPIQDicU5ufHE%3D",
    "sourceFilter": {
        "prefix":"Invoice_",
        "includeSubFolders":false
    },
    "useLabelFIle":true
}

GET API Response :-

{
    "modelInfo": {
        "modelId": "622bf015-6a89-45d5-a576-98e798b72775",
        "status": "invalid",
        "createdDateTime": "2022-12-06T00:10:38Z",
        "lastUpdatedDateTime": "2022-12-06T00:10:39Z"
    },
    "trainResult": {
        "averageModelAccuracy": 0.0,
        "errors": [
            {
                "code": "2100",
                "message": "Can't find any OCR files for training."
            }
        ]
    }
}

Evidence of labelled OCR files:-

enter image description here

Upendra Dama
  • 99
  • 1
  • 8

2 Answers2

0

Yes, The form recognizer is working on pre-trained models and that can recognize the key-value pairs, text, and tables from your documents and the table contents in the file uploaded as the input. Even though the file contains a large amount of text in paragraphs and table content in the middle or at any place, it will be recognized.

labeling OCR files refer this Link.

For more information refer this SO thread and GitHub link.

B. B. Naga Sai Vamsi
  • 2,386
  • 2
  • 3
  • 11
0

The error occurs because your Form Recognizer Studio is not able to access your blob storage. To Enable that

  1. Go to Blob Storage resource
  2. Select relevant containers
  3. Click on "Change Access Level"
  4. Select Anonymous read access for container & blobs option.
Abdul Rafay
  • 46
  • 4