search

clovaai / deep-text-recognition-benchmark

Text recognition (optical character recognition) with deep learning methods, ICCV 2019
Apache License 2.0
3.69k stars 1.08k forks source link

Fine tuning the EasyOCR #367

Open riccardoregnicoli opened 1 year ago

riccardoregnicoli commented 1 year ago

Hi! I wanted to ask if there is some tutorial I can follow to fine tune the easy ocr model... I have some images... (I presume I have to prepare the file with labels) but, in order to fine tune the model with my data (once I have my dataset prepared), which code should I use?

Thank you all, and I am sorry if this is a banal question

riccardoregnicoli commented 1 year ago

Hi everyone, I managed to prepare the dataset with labels and to fine tune I have used this modified version (https://github.com/JaidedAI/EasyOCR/blob/master/trainer/trainer.ipynb) I have used the model avalaible here(https://jaided.ai/easyocr/modelhub/) , the englishg2.pth

But I get best accuracy 6.63 Why is it so low? are there some parameters which I have to modify? Maybe I'm doing something wrong or it could be the dimension of images? Have I to prepare a dataset of images with a fixed size?

Thank you all!!