Spaces:
Sleeping
Sleeping
| import gradio as gr | |
| from mineru_vl_utils.mineru_client import MinerUClient | |
| from PIL import Image | |
| import fitz # PyMuPDF | |
| import os | |
| import torch | |
| # Init client | |
| model_path = "opendatalab/MinerU2.5-2509-1.2B" | |
| device = "cuda" if torch.cuda.is_available() else "cpu" | |
| print(f"⚡ Loading MinerU on device: {device}") | |
| client = MinerUClient(backend="transformers", model_path=model_path) | |
| def extract_from_file(file): | |
| try: | |
| print(f"📄 Processing file: {file.name}") | |
| ext = os.path.splitext(file.name)[-1].lower() | |
| images = [] | |
| if ext == ".pdf": | |
| doc = fitz.open(file.name) | |
| for i, page in enumerate(doc): | |
| pix = page.get_pixmap() | |
| img = Image.frombytes("RGB", [pix.width, pix.height], pix.samples) | |
| images.append(img) | |
| print(f"✅ Converted page {i+1} to image") | |
| else: | |
| images.append(Image.open(file.name)) | |
| results = [] | |
| for i, img in enumerate(images): | |
| print(f"🔍 Extracting text from page {i+1}") | |
| blocks = client.two_step_extract(img) | |
| if not blocks: | |
| print(f"⚠️ No blocks found on page {i+1}") | |
| results.append("[EMPTY PAGE]") | |
| continue | |
| text_blocks = [] | |
| for b in blocks: | |
| if hasattr(b, "text") and b.text: | |
| text_blocks.append(b.text) | |
| page_text = "\n".join(text_blocks) if text_blocks else "[NO TEXT FOUND]" | |
| results.append(page_text) | |
| print(f"✅ Page {i+1} extracted") | |
| return "\n\n--- PAGE ---\n\n".join(results) | |
| except Exception as e: | |
| print(f"❌ ERROR: {e}") | |
| return f"Error during extraction: {str(e)}" | |
| demo = gr.Interface( | |
| fn=extract_from_file, | |
| inputs=gr.File(type="filepath", label="Upload PDF or Image"), | |
| outputs=gr.Textbox(label="Extracted Text", lines=20), | |
| title="MinerU2.5 Document Extractor", | |
| description="Upload a PDF or Image to extract structured text using MinerU2.5." | |
| ) | |
| demo.launch() |