Upload 3 files

OCR-and-Keyword-Search-with-Qwen2-VL/README.md

@@ -0,0 +1,50 @@
+# Hindi English OCR with Keyword Search using Qwen2-VL
+
+![{60950997-B613-4DCA-A2F0-AF993237BDE1}](https://github.com/user-attachments/assets/3c742784-509b-4198-8e91-75d19b519f9f)
+
+This project implements a powerful OCR tool that utilizes the **Qwen2-VL** model for extracting text from images in Hindi and English. Users can upload images, extract the text, and optionally search for specific keywords within the extracted text.
+
+## Features
+
+- Upload images with text in Hindi or English.
+- Extracted text displayed in a user-friendly interface.
+- Optional keyword search functionality to highlight occurrences in the extracted text.
+
+## Requirements
+
+Before running the application, ensure you have the following Python packages installed:
+
+```bash
+pip install gradio transformers Pillow torch
+
+```
+Usage
+Clone the repository:
+```
+git clone https://github.com/Ashutosh0x/OCR-and-Keyword-Search-with-Qwen2-VL.git
+cd OCR-and-Keyword-Search-with-Qwen2-VL 
+```
+
+Install the required packages.
+
+Run the application:
+python app.py
+
+How It Works
+The application uses the Qwen2-VL model to perform OCR on uploaded images.
+Users can select a language for the OCR process and enter keywords for searching within the extracted text.
+The app provides instant feedback and displays both the extracted text and search results.
+Model
+This application utilizes the Qwen2-VL model, which is designed for visual-language tasks, enabling efficient text extraction and processing from images.
+
+Gradio for the easy-to-use interface.
+Transformers library for model handling.
+Pillow for image processing.
+
+Deploying on Hugging Face
+You can easily deploy this application on Hugging Face Spaces. Follow these steps:
+
+Go to Hugging Face Spaces.
+Create a new Space and choose the "Gradio" option.
+Upload your app.py file and any other necessary files (like requirements.txt).
+Once uploaded, Hugging Face will automatically build and run your application.

OCR-and-Keyword-Search-with-Qwen2-VL/app.py

@@ -0,0 +1,110 @@
+import gradio as gr
+from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
+from PIL import Image
+import torch
+import re
+
+# Load the Qwen2-VL model and processor
+model = Qwen2VLForConditionalGeneration.from_pretrained(
+    "Qwen/Qwen2-VL-2B-Instruct", torch_dtype="auto", device_map={"": "cpu"}
+)
+processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-2B-Instruct")
+
+# Function to perform OCR using Qwen2-VL
+def ocr_image(image, language):
+    try:
+        # Prepare the input format for Qwen2-VL
+        messages = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "image", "image": image},
+                    {"type": "text", "text": "Extract the text from this image."},
+                ],
+            }
+        ]
+
+        # Process the input for the model
+        text_input = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+        inputs = processor(images=image, text=[text_input], padding=True, return_tensors="pt").to("cpu")
+
+        # Generate text using the model
+        generated_ids = model.generate(**inputs, max_new_tokens=128)
+        extracted_text = processor.batch_decode(generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        
+        return extracted_text
+    except Exception as e:
+        return f"Error occurred during OCR: {str(e)}"
+
+# Function to perform keyword search within the extracted text
+def search_text(text, keyword):
+    if not text:
+        return "No text extracted."
+
+    # Search for keyword occurrences (case insensitive)
+    keyword_pattern = re.compile(re.escape(keyword), re.IGNORECASE)
+    matches = keyword_pattern.findall(text)
+
+    if matches:
+        # Highlight all matches in the text by wrapping them with ** for bold
+        highlighted_text = re.sub(keyword_pattern, f"**{keyword}**", text)
+        return highlighted_text
+    else:
+        return "Keyword not found in the text."
+
+# Function to handle both OCR and search functionality
+def process_image(image, language, keyword=""):
+    extracted_text = ocr_image(image, language)
+
+    if keyword:
+        result_text = search_text(extracted_text, keyword)
+    else:
+        result_text = extracted_text
+
+    return extracted_text, result_text
+
+# Gradio Interface
+def build_interface():
+    with gr.Blocks() as interface:
+        gr.Markdown("## Hindi & English OCR with Keyword Search using Qwen2-VL")
+        gr.Markdown("Upload an image with text in Hindi or English, extract the text, and optionally search for keywords within it.")
+
+        with gr.Row():
+            with gr.Column(scale=1):
+                image_input = gr.Image(type="pil", label="Upload an Image", elem_id="image-input", height=400)
+
+            with gr.Column(scale=1):
+                # Dropdown for selecting language (currently not used by Qwen2-VL, but kept for future expansion)
+                language_input = gr.Dropdown(label="Select Language for OCR", choices=["English", "Hindi", "Both"], value="Both", elem_id="lang-dropdown", interactive=True)
+
+                keyword_input = gr.Textbox(label="Enter Keyword to Search (Optional)", value="", placeholder="Enter keyword here", elem_id="keyword-input", interactive=True)
+
+        with gr.Row():
+            text_output = gr.Textbox(label="Extracted Text", interactive=False, placeholder="Extracted text will appear here.", elem_id="text-output", lines=10)
+            search_output = gr.Textbox(label="Search Results", interactive=False, placeholder="Search results will appear here.", elem_id="search-output", lines=10)
+
+        # Button to submit the image, language, and keyword
+        submit_btn = gr.Button("Process", elem_id="submit-btn")
+
+        # Markdown for feedback during processing
+        processing_message = gr.Markdown("Processing...", visible=False, elem_id="processing-message")
+
+        def on_submit(image, language, keyword):
+            if not image:
+                return "No image provided", "Please upload an image.", gr.update(visible=False)
+
+            # Display the processing message
+            extracted_text, result_text = process_image(image, language, keyword)
+            return extracted_text, result_text, gr.update(visible=False)
+
+        # Connect the button to the function with inputs and outputs
+        submit_btn.click(on_submit,
+                         inputs=[image_input, language_input, keyword_input],
+                         outputs=[text_output, search_output, processing_message])
+
+    return interface
+
+# Main function to launch the app
+if __name__ == "__main__":
+    interface = build_interface()
+    interface.launch(share=True, inbrowser=True, debug=True, height=900, width=1600)

OCR-and-Keyword-Search-with-Qwen2-VL/requirements.txt

@@ -0,0 +1,6 @@
+gradio
+transformers
+torch
+Pillow
+regex
+accelerate>=0.26.0