Update app.py

app.py

@@ -10,81 +10,81 @@ processor = AutoProcessor.from_pretrained("allenai/olmOCR-7B-0225-preview")
 model = AutoModelForCausalLM.from_pretrained("allenai/olmOCR-7B-0225-preview")
 
 def process_pdf(pdf_file):
- """
- Process the uploaded PDF file, extract text from each page, and generate HTML
- to display each page's image and text with copy buttons.
- """
- # Check if a PDF file was uploaded
- if pdf_file is None:
- return "<p>Please upload a PDF file.</p>"
- 
- # Convert PDF to images
- try:
- pages = convert_from_path(pdf_file.name)
- except Exception as e:
- return f"<p>Error converting PDF to images: {str(e)}</p>"
- 
- # Start building the HTML output
- html = '<div><button onclick="copyAll()" style="margin-bottom: 10px;">Copy All</button></div><div id="pages">'
- 
- # Process each page
- for i, page in enumerate(pages):
- # Convert the page image to base64 for embedding in HTML
- buffered = io.BytesIO()
- page.save(buffered, format="PNG")
- img_str = base64.b64encode(buffered.getvalue()).decode()
- img_data = f"data:image/png;base64,{img_str}"
- 
- # Extract text from the page using the OCR model
- try:
- inputs = processor(text="Extract the text from this image.", images=page, return_tensors="pt")
- outputs = model.generate(**inputs)
- text = processor.decode(outputs[0], skip_special_tokens=True)
- except Exception as e:
- text = f"Error extracting text: {str(e)}"
- 
- # Generate HTML for this page's section
- textarea_id = f"text{i+1}"
- html += f'''
- <div class="page" style="margin-bottom: 20px; border-bottom: 1px solid #ccc; padding-bottom: 20px;">
- <h3>Page {i+1}</h3>
- <div style="display: flex; align-items: flex-start;">
- <img src="{img_data}" alt="Page {i+1}" style="max-width: 300px; margin-right: 20px;">
- <div style="flex-grow: 1;">
- <textarea id="{textarea_id}" rows="10" style="width: 100%;">{text}</textarea>
- <button onclick="copyText('{textarea_id}')" style="margin-top: 5px;">Copy</button>
- </div>
- </div>
- </div>
- '''
- 
- # Close the pages div and add JavaScript for copy functionality
- html += '</div>'
- html += '''
- <script>
- function copyText(id) {
- var text = document.getElementById(id);
- text.select();
- document.execCommand("copy");
- }
- function copyAll() {
- var texts = document.querySelectorAll("#pages textarea");
- var allText = Array.from(texts).map(t => t.value).join("\\n\\n");
- navigator.clipboard.writeText(allText);
- }
- </script>
- '''
- return html
+    """
+    Process the uploaded PDF file, extract text from each page, and generate HTML
+    to display each page's image and text with copy buttons.
+    """
+    # Check if a PDF file was uploaded
+    if pdf_file is None:
+        return "<p>Please upload a PDF file.</p>"
+    
+    # Convert PDF to images
+    try:
+        pages = convert_from_path(pdf_file.name)
+    except Exception as e:
+        return f"<p>Error converting PDF to images: {str(e)}</p>"
+    
+    # Start building the HTML output
+    html = '<div><button onclick="copyAll()" style="margin-bottom: 10px;">Copy All</button></div><div id="pages">'
+    
+    # Process each page
+    for i, page in enumerate(pages):
+        # Convert the page image to base64 for embedding in HTML
+        buffered = io.BytesIO()
+        page.save(buffered, format="PNG")
+        img_str = base64.b64encode(buffered.getvalue()).decode()
+        img_data = f"data:image/png;base64,{img_str}"
+        
+        # Extract text from the page using the OCR model
+        try:
+            inputs = processor(text="Extract the text from this image.", images=page, return_tensors="pt")
+            outputs = model.generate(**inputs)
+            text = processor.decode(outputs[0], skip_special_tokens=True)
+        except Exception as e:
+            text = f"Error extracting text: {str(e)}"
+        
+        # Generate HTML for this page's section
+        textarea_id = f"text{i+1}"
+        html += f'''
+        <div class="page" style="margin-bottom: 20px; border-bottom: 1px solid #ccc; padding-bottom: 20px;">
+            <h3>Page {i+1}</h3>
+            <div style="display: flex; align-items: flex-start;">
+                <img src="{img_data}" alt="Page {i+1}" style="max-width: 300px; margin-right: 20px;">
+                <div style="flex-grow: 1;">
+                    <textarea id="{textarea_id}" rows="10" style="width: 100%;">{text}</textarea>
+                    <button onclick="copyText('{textarea_id}')" style="margin-top: 5px;">Copy</button>
+                </div>
+            </div>
+        </div>
+        '''
+    
+    # Close the pages div and add JavaScript for copy functionality
+    html += '</div>'
+    html += '''
+    <script>
+    function copyText(id) {
+        var text = document.getElementById(id);
+        text.select();
+        document.execCommand("copy");
+    }
+    function copyAll() {
+        var texts = document.querySelectorAll("#pages textarea");
+        var allText = Array.from(texts).map(t => t.value).join("\\n\\n");
+        navigator.clipboard.writeText(allText);
+    }
+    </script>
+    '''
+    return html
 
 # Define the Gradio interface
 with gr.Blocks(title="PDF Text Extractor") as demo:
- gr.Markdown("# PDF Text Extractor")
- gr.Markdown("Upload a PDF file and click 'Extract Text' to see each page's image and extracted text.")
- with gr.Row():
- pdf_input = gr.File(label="Upload PDF", file_types=[".pdf"])
- submit_btn = gr.Button("Extract Text")
- output_html = gr.HTML()
- submit_btn.click(fn=process_pdf, inputs=pdf_input, outputs=output_html)
+    gr.Markdown("# PDF Text Extractor")
+    gr.Markdown("Upload a PDF file and click 'Extract Text' to see each page's image and extracted text.")
+    with gr.Row():
+        pdf_input = gr.File(label="Upload PDF", file_types=[".pdf"])
+        submit_btn = gr.Button("Extract Text")
+    output_html = gr.HTML()
+    submit_btn.click(fn=process_pdf, inputs=pdf_input, outputs=output_html)
 
 # Launch the interface
 demo.launch()