app.py

import gradio as gr
from transformers import  VisionEncoderDecoderModel, AutoImageProcessor, BertTokenizerFast
import requests
from PIL import Image

urls = ['https://huggingface.co/spaces/sivan22/TrOCR-handwritten-hebrew/resolve/main/article_1_page_10line_1.png', 'https://huggingface.co/spaces/sivan22/TrOCR-handwritten-hebrew/resolve/main/article_1_page_10line_10.png',
        'https://huggingface.co/spaces/sivan22/TrOCR-handwritten-hebrew/resolve/main/article_1_page_10line_11.png']
for idx, url in enumerate(urls):
  image = Image.open(requests.get(url, stream=True).raw)
  image.save(f"image_{idx}.png")

from transformers import BertTokenizer, BasicTokenizer
from transformers.tokenization_utils import _is_punctuation

class OurBasicTokenizer(BasicTokenizer):
    def _run_split_on_punc(self, text, never_split=None):
        """Splits punctuation on a piece of text."""
        if text in self.never_split or (never_split and text in never_split):
            return [text]
        chars = list(text)
        i = 0
        start_new_word = True
        output = []
        while i < len(chars):
            char = chars[i]
            if _is_punctuation(char) and char != "'" and not (char == '"' and i + 1 < len(chars) and not _is_punctuation(chars[i + 1])):
                output.append([char])
                start_new_word = True
            else:
                if start_new_word:
                    output.append([])
                start_new_word = False
                output[-1].append(char)
            i += 1

        return ["".join(x) for x in output]


def RabbinicTokenizer(tok):
    tok.basic_tokenizer = OurBasicTokenizer(tok.basic_tokenizer.do_lower_case, tok.basic_tokenizer.never_split)
    return tok


image_processor = AutoImageProcessor.from_pretrained("microsoft/swinv2-tiny-patch4-window8-256")
tokenizer = RabbinicTokenizer(BertTokenizer.from_pretrained("sivan22/BEREL"))
model = VisionEncoderDecoderModel.from_pretrained("sivan22/ABBA-HTR")


def process_image(image):
    # prepare image
    pixel_values = image_processor(image, return_tensors="pt").pixel_values

    # generate (no beam search)
    generated_ids = model.generate(pixel_values)

    # decode
    generated_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)

    return generated_text

title = "הדגמה: פענוח כתב יד באמצעות בינה מלאכותית"
description = "על בסיס טכנולוגיית trOCR"
article = "<p style='text-align: center'>sivan22</p>"
examples =[["article_1_page_10line_1.png"], ["article_1_page_10line_10.png"], ["article_1_page_10line_11.png"]]

#css = """.output_image, .input_image {height: 600px !important}"""

iface = gr.Interface(fn=process_image, 
                     inputs=gr.inputs.Image(type="pil"), 
                     outputs=gr.outputs.Textbox(),
                     title=title,
                     description=description,
                     article=article,
                     examples=examples)
iface.launch(debug=True)