app.py

import gradio as gr
from PIL import Image, ImageEnhance, ImageOps
import torch
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
from datetime import datetime
import pytz
import re

# Load model and processor
processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-stage1")
model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-base-stage1")

# Enhance image before OCR
def enhance_image(image):
    image = image.convert("L")  # grayscale
    image = ImageOps.invert(image)  # invert colors
    image = ImageEnhance.Contrast(image).enhance(2)  # boost contrast
    image = ImageEnhance.Sharpness(image).enhance(2)  # sharpen
    image = image.resize((image.width * 2, image.height * 2))  # enlarge
    return image

# Extract weight
def detect_weight(image):
    try:
        processed_image = enhance_image(image)
        pixel_values = processor(images=processed_image, return_tensors="pt").pixel_values
        generated_ids = model.generate(pixel_values)
        generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

        # Extract number
        match = re.search(r"(\d{1,4}(?:\.\d{1,2})?)", generated_text)
        weight = match.group(1) if match else "Not detected"

        # Timestamp
        ist = pytz.timezone('Asia/Kolkata')
        current_time = datetime.now(ist).strftime("%Y-%m-%d %H:%M:%S")

        return f"Weight: {weight} kg\nCaptured At: {current_time}", image
    except Exception as e:
        return f"Error: {str(e)}", image

# Gradio UI
interface = gr.Interface(
    fn=detect_weight,
    inputs=gr.Image(type="pil", label="Upload or Capture Image"),
    outputs=[gr.Textbox(label="Weight Info"), gr.Image(label="Snapshot")],
    title="⚖️ Auto Weight Detector (No Tesseract)",
    description="Detects weight from digital scale image using Hugging Face TrOCR."
)

interface.launch()