AWS Textract thinks non-driver identification cards are driver licenses

2

When calling AnalyzeId on a non-driver ID, the response from textract indicates that the ID_TYPE is DRIVER LICENSE FRONT

Here's a fake document I pulled from the internet: CA Non driver ID

Here's the response from textract (I've omitted the Blocks to make this question more concise). Note that ID_TYPE is DRIVER LICENSE FRONT with high confidence!

Is this a bug?

{
  "IdentityDocuments": [
    {
      "DocumentIndex": 1,
      "IdentityDocumentFields": [
        {
          "Type": {
            "Text": "FIRST_NAME"
          },
          "ValueDetection": {
            "Text": "SEIJI",
            "Confidence": 97.91222381591797
          }
        },
        {
          "Type": {
            "Text": "LAST_NAME"
          },
          "ValueDetection": {
            "Text": "SAMPLE",
            "Confidence": 74.50897216796875
          }
        },
        {
          "Type": {
            "Text": "MIDDLE_NAME"
          },
          "ValueDetection": {
            "Text": "",
            "Confidence": 99.19636535644531
          }
        },
        {
          "Type": {
            "Text": "SUFFIX"
          },
          "ValueDetection": {
            "Text": "",
            "Confidence": 99.15950012207031
          }
        },
        {
          "Type": {
            "Text": "CITY_IN_ADDRESS"
          },
          "ValueDetection": {
            "Text": "SACRAMENTO",
            "Confidence": 95.51748657226562
          }
        },
        {
          "Type": {
            "Text": "ZIP_CODE_IN_ADDRESS"
          },
          "ValueDetection": {
            "Text": "95818",
            "Confidence": 97.53475952148438
          }
        },
        {
          "Type": {
            "Text": "STATE_IN_ADDRESS"
          },
          "ValueDetection": {
            "Text": "CA",
            "Confidence": 98.85255432128906
          }
        },
        {
          "Type": {
            "Text": "STATE_NAME"
          },
          "ValueDetection": {
            "Text": "CALIFORNIA",
            "Confidence": 96.00836944580078
          }
        },
        {
          "Type": {
            "Text": "DOCUMENT_NUMBER"
          },
          "ValueDetection": {
            "Text": "I1234561",
            "Confidence": 96.42362976074219
          }
        },
        {
          "Type": {
            "Text": "EXPIRATION_DATE"
          },
          "ValueDetection": {
            "Text": "08/31/2015",
            "NormalizedValue": {
              "Value": "2015-08-31T00:00:00",
              "ValueType": "Date"
            },
            "Confidence": 95.50371551513672
          }
        },
        {
          "Type": {
            "Text": "DATE_OF_BIRTH"
          },
          "ValueDetection": {
            "Text": "08/31/1977",
            "NormalizedValue": {
              "Value": "1977-08-31T00:00:00",
              "ValueType": "Date"
            },
            "Confidence": 96.58924865722656
          }
        },
        {
          "Type": {
            "Text": "DATE_OF_ISSUE"
          },
          "ValueDetection": {
            "Text": "09/30/2010",
            "NormalizedValue": {
              "Value": "2010-09-30T00:00:00",
              "ValueType": "Date"
            },
            "Confidence": 96.66231536865234
          }
        },
        {
          "Type": {
            "Text": "ID_TYPE"
          },
          "ValueDetection": {
            "Text": "DRIVER LICENSE FRONT",
            "Confidence": 97.12804412841797
          }
        },
        {
          "Type": {
            "Text": "ENDORSEMENTS"
          },
          "ValueDetection": {
            "Text": "",
            "Confidence": 99.18586730957031
          }
        },
        {
          "Type": {
            "Text": "VETERAN"
          },
          "ValueDetection": {
            "Text": "",
            "Confidence": 99.18095397949219
          }
        },
        {
          "Type": {
            "Text": "RESTRICTIONS"
          },
          "ValueDetection": {
            "Text": "",
            "Confidence": 99.15245056152344
          }
        },
        {
          "Type": {
            "Text": "CLASS"
          },
          "ValueDetection": {
            "Text": "",
            "Confidence": 99.16776275634766
          }
        },
        {
          "Type": {
            "Text": "ADDRESS"
          },
          "ValueDetection": {
            "Text": "2570 24TH STREET",
            "Confidence": 96.71549224853516
          }
        },
        {
          "Type": {
            "Text": "COUNTY"
          },
          "ValueDetection": {
            "Text": "",
            "Confidence": 99.175537109375
          }
        },
        {
          "Type": {
            "Text": "PLACE_OF_BIRTH"
          },
          "ValueDetection": {
            "Text": "",
            "Confidence": 99.1697998046875
          }
        },
        {
          "Type": {
            "Text": "MRZ_CODE"
          },
          "ValueDetection": {
            "Text": "",
            "Confidence": 97.52288055419922
          }
        }
      ],
      "Blocks": [
        ...
      ]
    }
  ],
  "DocumentMetadata": {
    "Pages": 1
  },
  "AnalyzeIDModelVersion": "1.0"
}
asked a year ago364 views
2 Answers
0

I have replicated the issue on my end. Probably the model has miss predicted this ID type. Please open a case to inform with the findings.

AWS
answered a year ago
  • Thanks for the response and replication! Will open a case!

0

From my replication, it seems that the ID document provided has been identified as a Driver's License when using the AnalyzeID API. Please note, Amazon Textract is built on an orchestration of Machine Learning technology which gets consistently improved to a support variety of documents. While the internal Textract team do their best to make sure that the results and their accuracy improve over a period of time, there are minimal occasions where the results aren't as accurate. This is a side-effect of the underlying technology. As such, can you please create a case to the AWS Premium Support Textract team for them to have a look into this further.

answered a year ago
  • Thanks for responding and replicating! I'll create a case, and will post updates if/when it gets resolved

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions