A CRNN model for Persian OCR. This model is based on a simple CNN + LSTM architecture inspired by this paper.
This is a successor model to our previous model hezarai/crnn-base-fa-v1. The improvements include:
- 5X larger dataset
- Change input image size from 64x256 to 32x384
- Increase max output length from 64 to 96 (Max length of the samples in the dataset was 48 to handle CTC loss issues)
- Support numbers and special characters (see id2label in
model_config.yaml) - Auto-handling of LTR characters like digits in between the text
Note that this model is only optimized for printed/scanned documents and works best on texts with a length of up to 50-ish characters. (For an end-to-end OCR pipeline, use a text detector model first like https://huggingface.co/hezarai/CRAFT to extract text boxes preferrably in word-level and then use this model), but it can be used to be fine-tuned on other domains like license plate or handwritten texts.
Usage
pip install hezar
from hezar.models import Model
crnn = Model.load("hezarai/crnn-base-fa-v2")
texts = crnn.predict(["sample_image.jpg"])
print(texts)