app.py

import os
import re
import torch
import traceback
import gradio as gr
from transformers import DonutProcessor, VisionEncoderDecoderModel

# ─── 모델 로딩 ─────────────────────────────────────────────────────────
MODEL_NAME = "naver-clova-ix/donut-base-finetuned-cord-v2"
processor = DonutProcessor.from_pretrained(MODEL_NAME)
model = VisionEncoderDecoderModel.from_pretrained(MODEL_NAME)

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

# ─── OCR 함수 ──────────────────────────────────────────────────────────
def ocr_donut(image):
    try:
        if image is None:
            return {"error": "No image provided."}
        task_prompt = "<s_cord-v2>"
        decoder_input_ids = processor.tokenizer(
            task_prompt, add_special_tokens=False, return_tensors="pt"
        ).input_ids.to(device)
        pixel_values = processor(image.convert("RGB"), return_tensors="pt").pixel_values.to(device)

        outputs = model.generate(
            pixel_values,
            decoder_input_ids=decoder_input_ids,
            max_length=model.config.decoder.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,
        )

        seq = processor.batch_decode(outputs.sequences)[0]
        seq = seq.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")
        seq = re.sub(r"<.*?>", "", seq, count=1).strip()
        return {"result": processor.token2json(seq)}

    except Exception:
        tb = traceback.format_exc()
        print(tb)
        return {"error": tb}

# ─── CSS 스타일링 ────────────────────────────────────────────────────
custom_css = """
body { background: #f0f2f5; font-family: 'Segoe UI', Tahoma, sans-serif; }
.gradio-container { max-width: 900px; margin: 40px auto; padding: 20px; }
.header { text-align: center; margin-bottom: 30px; }
.header h1 { font-size: 2.8rem; color: #333; margin: 0; }
.header p { color: #666; margin-top: 8px; }

.input-box, .output-box {
    background: #fff;
    border-radius: 8px;
    box-shadow: 0 2px 8px rgba(0,0,0,0.1);
    padding: 20px;
}
.input-box { margin-right: 10px; }
.output-box { margin-left: 10px; }

.gr-button {
    background: #5a8dee !important;
    color: #fff !important;
    border-radius: 6px !important;
    padding: 10px 20px !important;
    font-size: 1rem !important;
    margin-top: 10px !important;
    transition: background 0.2s ease;
}
.gr-button:hover { background: #3f6fcc !important; }

.footer {
    text-align: center;
    margin-top: 30px;
    color: #999;
    font-size: 0.85rem;
}
"""

# ─── Blocks 레이아웃 ──────────────────────────────────────────────────
with gr.Blocks(css=custom_css, title="Donut OCR App") as demo:
    # 헤더
    gr.HTML(
        """
        <div class="header">
          <h1>📄 Donut OCR</h1>
          <p>Industrial AI Engineering Week 8 Assignment</p>        
        </div>
        """
    )

    # 입력/출력 영역
    with gr.Row():
        with gr.Column(elem_classes="input-box"):
            image_input = gr.Image(type="pil", label="Upload Document Image")
            run_btn = gr.Button("Run OCR", elem_id="run-btn")
        with gr.Column(elem_classes="output-box"):
            result_box = gr.JSON(label="Output")

    # 버튼 클릭 연결
    run_btn.click(fn=ocr_donut, inputs=image_input, outputs=result_box)

    # 푸터
    gr.HTML(
        """
        <div class="footer">
          <p>Powered by Naver Clova Donut</p>
        </div>
        """
    )

# Spaces 실행
demo.launch(
    server_name="0.0.0.0",
    server_port=int(os.environ.get("PORT", 7860)),
    debug=True
)