Power Platform Community Forum Thread Details

Could someone please assist me in troubleshooting an issue with my Power Automate Desktop flow?

I've utilized the "OCR" action to extract text from an image using Microsoft Cognitive Services with an Azure subscription key.

However, despite providing the correct image as input[screenshot-1], the OCR action is yielding inaccurate results[screenshot-2], as shown in the JSON response below:
Image used to send to the OCR and curl command

screenshot-1

Output - OCR action

screenshot -2

{
"language": "en",
"textAngle": 0.0349065850398872,
"orientation": "Down",
"regions": [
{
"boundingBox": "50,108,118,176",
"lines": [
{
"boundingBox": "56,108,49,15",
"words": [
{
"boundingBox": "56,108,49,15",
"text": "(paved)"
}
]
},
{
"boundingBox": "55,128,113,16",
"words": [
{
"boundingBox": "55,128,49,15",
"text": "(paved)"
},
{
"boundingBox": "108,130,60,14",
"text": "Åuq51ua_J04"
}
]
},
{
"boundingBox": "54,148,113,16",
"words": [
{
"boundingBox": "54,148,49,15",
"text": "(paved)"
},
{
"boundingBox": "107,150,60,14",
"text": "ÅAq5.uuos"
}
]
},
{
"boundingBox": "54,168,112,16",
"words": [
{
"boundingBox": "54,168,49,15",
"text": "(paved)"
},
{
"boundingBox": "107,170,59,14",
"text": "Kug6'uuos"
}
]
},
{
"boundingBox": "53,188,49,15",
"words": [
{
"boundingBox": "53,188,49,15",
"text": "(paved)"
}
]
},
{
"boundingBox": "52,208,113,16",
"words": [
{
"boundingBox": "52,208,49,15",
"text": "(paved)"
},
{
"boundingBox": "105,210,60,14",
"text": ",4nq5.uu0d"
}
]
},
{
"boundingBox": "52,228,112,16",
"words": [
{
"boundingBox": "52,228,49,15",
"text": "(paved)"
},
{
"boundingBox": "104,230,60,14",
"text": ",qu61uuos"
}
]
},
{
"boundingBox": "51,248,112,16",
"words": [
{
"boundingBox": "51,248,49,15",
"text": "(paved)"
},
{
"boundingBox": "104,250,59,14",
"text": ",qaq51uuos"
}
]
},
{
"boundingBox": "50,268,113,16",
"words": [
{
"boundingBox": "50,268,49,15",
"text": "(paved)"
},
{
"boundingBox": "103,270,60,14",
"text": ",4gq51uuos"
}
]
}
]
}
]
}

Interestingly, when I used the curl method (REST API) for, the OCR operation worked perfectly and provided the expected output, as depicted in the attached screenshot.
Note: Here, I have added some filter commands to get the expected output from the json output of curl command.
Output - curl command
What could be causing this discrepancy with the "OCR" action in Power Automate Desktop?