File size: 1,322 Bytes
5c76600 820b26f 9492734 820b26f 9492734 9160227 be357a5 9160227 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 |
---
license: apache-2.0
language:
- fa
library_name: hezar
tags:
- hezar
- image-to-text
pipeline_tag: image-to-text
---
A CRNN model for Persian OCR. This model is based on a simple CNN + LSTM architecture inspired by [this paper](https://arxiv.org/abs/1507.05717).
This is a successor model to our previous model [hezarai/crnn-base-fa-64x256](https://huggingface.co./hezarai/crnn-base-fa-64x256).
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 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
```
```python
from hezar.models import Model
crnn = Model.load("hezarai/crnn-fa-printed-96-long")
texts = crnn.predict(["sample_image.jpg"])
print(texts)
``` |