How it works
VLens will provide on prem ENID finetuning tool that you would need to input only 1 parameter for it to start your enhancement round right away !
Input the path includes data that you are 100% confident of its correctness, and start finetuning your OCR models.
Input data is automatically split into 2 parts, 1st part is 85% of the data and it is used in background finetuning, 2nd part is 15% of the data and it would be used in measuring the accuracies of models.
Once training process for new models is finished, accuracies measurement process would start, the 15% testing data would be sent to both the current deployed models & new trained models, and accuracies per field should be the result for both old & new models so that the tool would decide which field is enhanced and which is not.
New trained models that have better accuracies would be deployed automatically and customer would be notified of which fields models are changed and the accuracies uplift.
Last updated