@Eltee The computer vision API Read API is asynchronous where the request is processed after submission and the result with the entire content of the image is returned in the corresponding operation location. This does not provide an option to filter the results as part of the original request.
To ensure the required response you will have to write some client-side processing logic to read the confidence score of the response or compare the result with the required values i.e numeric values.
As per my experience using the confidence score is advisable if you have an input image with numeric values and if one of the responses detects a non-numeric value with a low confidence score, you could use a threshold limit for confidence score in your application and if anything, below the threshold is returned it can be discarded or you can further compare the same with numeric values and then reject the response.
I hope this helps!!
If an answer is helpful, please click on or upvote
which might help other community members reading this thread.