app.py

import requests
from PIL import Image
import streamlit as st
# Load model directly
from transformers import AutoTokenizer, AutoModelForImageTextToText
from transformers import TrOCRProcessor, VisionEncoderDecoderModel

tokenizer = AutoTokenizer.from_pretrained("microsoft/trocr-large-printed")
model = AutoModelForImageTextToText.from_pretrained("microsoft/trocr-large-printed")

st.title("Duh!")

# load image from the IAM dataset
url = "https://parivahan.gov.in/rcdlstatus/DispplayCaptcha?txtp_cd=1&bkgp_cd=2&noise_cd=2&gimp_cd=3&txtp_length=5&pfdrid_c=true?1429026471&pfdrid_c=true"
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")

col1, col2 = st.columns(2)
processor = TrOCRProcessor.from_pretrained('microsoft/trocr-large-printed')
model = VisionEncoderDecoderModel.from_pretrained('microsoft/trocr-large-printed')
pixel_values = processor(images=image, return_tensors="pt").pixel_values

generated_ids = model.generate(pixel_values)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]


col1.image(image, use_column_width=True)
col2.subheader(f"Detected Text: {generated_text}")