import gradio as gr
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
from PIL import Image
import numpy as np
import torch

# Load TrOCR model and processor
processor = TrOCRProcessor.from_pretrained("microsoft/trocr-large-handwritten")
model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-large-handwritten")

# Function to extract text from handwritten images
def extract_text(image):
    try:
        # Convert image to RGB if needed
        if isinstance(image, np.ndarray):  
            if len(image.shape) == 2:  # If grayscale (H, W), convert to RGB
                image = np.stack([image] * 3, axis=-1)
            image = Image.fromarray(image)
        else:
            image = Image.open(image).convert("RGB")

        # Preprocessing (convert to grayscale for better OCR)
        image = image.convert("L")  
        image = image.resize((640, 640))  

        # Process image
        pixel_values = processor(images=image, return_tensors="pt").pixel_values
        generated_ids = model.generate(pixel_values)
        extracted_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

        return extracted_text if extracted_text.strip() else "No text detected."
    
    except Exception as e:
        return f"Error: {str(e)}"

# Gradio Interface
iface = gr.Interface(
    fn=extract_text,
    inputs="image",
    outputs="text",
    title="Handwritten OCR Extractor",
    description="Upload a handwritten image to extract text.",
)

# Launch the app
iface.launch()