UiPath Document Understanding + Google Cloud Vision Update

[Eileen Wilkerson]

My team was using the google cloud vision api as the ocr engine with document understanding in an AP invoicing use case. Back in May there was an update to the Google API and we are now experiencing some issues - i.e. the invoice number needs to be read as “RI-1234” and now it reads as “RI1234” omitting the “-“ which is required. These invoices are high quality pdfs and have been trained for in AI Center.

The legacy version expires on August 20. Has anyone else experienced issues with this API and has any other ideas for a fix?

Thanks!

[Eduardo Ortiz Caraveo]

Could you try and use “WEB_DETECTION” as the example below;

{

  "requests": [

    {

      "features": [

        {

          "type": "WEB_DETECTION"

        }

      ],

      "image": {

        "source": {

          "gcsImageUri": "gs://cloud-samples-data/vision/web/carnaval.jpeg"

        }

      },

      "imageContext": {

        "webDetectionParams": {

          "includeGeoResults": true

        }

      }

    }

  ]

}

Just instead of the image, make sure to specify that you are using a document.