from transformers import M2M100ForConditionalGeneration, M2M100Tokenizer

import pytesseract as tsr
from PIL import Image

import sys, os
import gradio as gr

tsr.pytesseract.tesseract_cmd = r'/usr/bin/tesseract'

model = M2M100ForConditionalGeneration.from_pretrained("facebook/m2m100_418M")
tokenizer = M2M100Tokenizer.from_pretrained("facebook/m2m100_418M")

def extractAndTranslate(image):
    # Extract Text
    extractedText = tsr.image_to_string(image, lang='eng+hin')
    extractedTextFormatted = ' '.join(extractedText.split('\n'))
    
    # Translate 
    tokenizer.src_lang = "en"
    encodedText = tokenizer(extractedTextFormatted, return_tensors="pt")
    generatedTokens = model.generate(**encodedText, forced_bos_token_id=tokenizer.get_lang_id("hi"))
    
    return tokenizer.batch_decode(generatedTokens, skip_special_tokens=True)[0]

demoApp = gr.Interface(extractAndTranslate, "image", "text")

demoApp.launch()