0

We are testing Azure Computer Vision (POC) when I tested using Azure ML Notebook the computer image I chosen is not rendering correctly. I tested with Azure vision studio https://portal.vision.cognitive.azure.com/demo/extract-text-from-images

as well the results from Azure ML Python SDK is matching what I get from Cognitive vision studio.

https://i.postimg.cc/Hx5W2V4w/Picture1.png is the image I'm testing and results coming as

Results DE316538430 4531 2.9m .9,6 ic SUPER 70 HEAVY

Any suggestions or recommendations? so that instead of DE31 I can get DFSU in the first line and instead of 4531 I can get 45G1.

Nathan
  • 35
  • 10

2 Answers2

0

Azure Computer Vision Studio is working on the internal API calls and some of the images which we try to upload may not be clear. It is a random error occurred for the current scenario.

Reproduced the same scenario with another image as well and even the small text which was upload extracted perfectly.

Reason wrong text: The font style may not be understandable to the API from image which we are running without any subscription and endpoint. Sharing the reproduced sample image even with small sized text.

enter image description here

The below image is the input from the question.

enter image description here

The font style may be the issue as in the image mentioned above is having the standard font family. This is a rare and random error. Which cannot be occurred when we started using python code with subscription and end point keys. For reference of code block, check out

Sairam Tadepalli
  • 1,563
  • 1
  • 3
  • 11
0

I created support request they acknowledged it's a bug. Also the image works fine if it's horizontal when it's vertical it has issues. Working with MSFT.

Nathan
  • 35
  • 10