Rob hi,
when u train a custom model, the metrics exist but they're tucked away in the training result. u gotta use the api to fetch them. after training, make a get request to the 'get operation details' endpoint. look for the 'analyzeResult' field it holds the juicy stuff like precision, recall, and f1.
a quick curl example to grab the training result
curl -X GET "https://{your-endpoint}/formrecognizer/documentModels/{modelId}/operations/{operationId}?api-version=2023-07-31" -H "Ocp-Apim-Subscription-Key: {your-key}"
also, if u're using python, the sdk wraps this nicely. just call get_operation_details
and dig into the response.
try logging the entire api response, u might spot them hiding in a nested field.
this might help in other tools too )) when apis don’t surface metrics upfront, the raw response often has gold buried inside. worth looking into the http traffic while training, tools like fiddler or postman can save u hours.
and hey, if u're comparing models, keep an eye on the 'confidence' scores in the analyze result. they’re not full metrics, but they hint at performance.
good luck with your models! hope this unblocks u ))
Alex
and "yes" if you would follow me at Q&A - personaly thx.
P.S. If my answer help to you, please Accept my answer