Open onlyfocusman opened 1 year ago
You're going to need to do some retraining with your own dataset @onlyfocusman . This guide shows how to retrain the model "from end to end:" https://keras-ocr.readthedocs.io/en/latest/examples/end_to_end_training.html
The guide assumes you have a dataset procured and prepped for training already (or that you're using the provided one). So if you don't have the data ready, then that's obviously the first step.
I've had pretty amazing success just following that guide, but that was with ASCII character sets. Hopefully someone who has worked with non-ASCII characters can provide some insight or tips/tricks/pitfalls when retraining with other characters.
Good luck!
I tried english and korean but Actual and Predicted value is not matched ... Is it possible Match actual value and predicted values ? with Need train Korean? if so, How can I do traing? I wish Korean prediction value :) I attached image thank you in advance.^^![111](https://user-images.githubusercontent.com/32046741/234807244-7641fafd-58bf-43fa-b10f-f5ac8acd6edb.jpg)