app.py

import re

from transformers import DonutProcessor, VisionEncoderDecoderModel
from datasets import load_dataset
import torch
from PIL import Image
import numpy as np
import streamlit as st

processor = DonutProcessor.from_pretrained("naver-clova-ix/donut-base-finetuned-rvlcdip")
model = VisionEncoderDecoderModel.from_pretrained("naver-clova-ix/donut-base-finetuned-rvlcdip")

device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)

#image = Image.open(r"C:\Invoices\Sample Invoices\sample invoice 1.tif")
#image = image.convert("RGB")
#print(np.array(image).shape)


st.title("Classify Document Image")

file_name = st.file_uploader("Upload a candidate image")

if file_name is not None:
    col1, col2, col3 = st.columns(3)

    image = Image.open(file_name)
    image = image.convert("RGB")

    # load document image
    #dataset = load_dataset("hf-internal-testing/example-documents", split="test")
    #image = dataset[2]["image"]


    task_prompt = "<s_rvlcdip>"
    decoder_input_ids = processor.tokenizer(task_prompt, add_special_tokens=False, return_tensors="pt").input_ids

    pixel_values = processor(image, return_tensors="pt").pixel_values

    outputs = model.generate(
        pixel_values.to(device),
        decoder_input_ids=decoder_input_ids.to(device),
        max_length=model.decoder.config.max_position_embeddings,
        pad_token_id=processor.tokenizer.pad_token_id,
        eos_token_id=processor.tokenizer.eos_token_id,
        use_cache=True,
        bad_words_ids=[[processor.tokenizer.unk_token_id]],
        return_dict_in_generate=True,
    )

    sequence = processor.batch_decode(outputs.sequences)[0]
    sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")
    sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()  # remove first task start token
    print(processor.token2json(sequence))
    
    col1.image(image, use_column_width=True)

    col2.header("Results")
    col2.subheader(processor.token2json(sequence))
    
    processor_ext = DonutProcessor.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2")
    model_ext = VisionEncoderDecoderModel.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2")

    device = "cuda" if torch.cuda.is_available() else "cpu"
    model_ext.to(device)

    # prepare decoder inputs
    task_prompt = "<s_cord-v2>"
    decoder_input_ids = processor_ext.tokenizer(task_prompt, add_special_tokens=False, return_tensors="pt").input_ids

    pixel_values = processor_ext(image, return_tensors="pt").pixel_values

    outputs = model_ext.generate(
        pixel_values.to(device),
        decoder_input_ids=decoder_input_ids.to(device),
        max_length=model_ext.decoder.config.max_position_embeddings,
        pad_token_id=processor_ext.tokenizer.pad_token_id,
        eos_token_id=processor_ext.tokenizer.eos_token_id,
        use_cache=True,
        bad_words_ids=[[processor_ext.tokenizer.unk_token_id]],
        return_dict_in_generate=True,
    )

    sequence = processor_ext.batch_decode(outputs.sequences)[0]
    sequence = sequence.replace(processor_ext.tokenizer.eos_token, "").replace(processor_ext.tokenizer.pad_token, "")
    sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()  # remove first task start token
    print(processor_ext.token2json(sequence))
    col3.header("Features")
    col3.subheader(processor_ext.token2json(sequence))