app.py

import gradio as gr
import numpy as np
import cv2
import os
from ultralytics import YOLO
from PIL import Image
from src.table_ocr import TableEx   
from src.datum_ocr import DatumOCR
from src.categories import CATEGORIES as categories, generate_colors

os.environ['TESSDATA_PREFIX'] = './tessdata'

def load_model():
    """Load the custom YOLO model using Ultralytics"""
    # Load the model using the Ultralytics YOLO class
    model = YOLO('src/yoloCADex.pt')
    model.to('cpu')
    return model


def process_image(image):
    """Process the uploaded image with the YOLO model and return the results"""
    
    # Check if image is valid
    if image is None or not isinstance(image, Image.Image):
        return None, {"error": "Invalid image input"}, None, None, None
    
    model = load_model()
    category_colors = generate_colors(len(categories))
    img_array = np.array(image)                 # Convert to format expected by the model
    table_extractor = TableEx()                 # Initialize TableEx for table extraction
    date_extractor = DatumOCR()                 # Initialize DatumOCR for OCR
    results = model(img_array)                  # Run inference with CPU specified
    img_with_boxes = img_array.copy()           # Create a copy of the image for drawing
    detections = []                             # Initialize results table
    
    table_data = []                             # Storage for extracted table data and images
    table_images = []
    
    gdnt_data = []                             # Storage for extracted table data and images
    surface_data = []


    # Process results
    for result in results:
        boxes = result.boxes
        for box in boxes:
            # Get box coordinates
            x1, y1, x2, y2 = map(int, box.xyxy[0])
            
            # Get confidence and class
            conf = float(box.conf[0])
            cls_id = int(box.cls[0])
            
            if cls_id < len(categories):
                cls_name = categories[cls_id]
                color = category_colors[cls_id]
                
                if cls_name == "table":
                    table_region, extracted_info = table_extractor.extract_table_data(img_array, x1, y1, x2, y2)
                    if table_region is not None:
                        table_images.append(table_region)
                    if extracted_info is not None:
                        table_data.append(extracted_info)
            else:
                cls_name = f"Unknown ({cls_id})"
                color = (255, 255, 255)  # White for unknown categories

            label = f"{cls_name} {conf:.2f}"
            
            # Draw rectangle with category-specific color
            cv2.rectangle(img_with_boxes, (x1, y1), (x2, y2), color, 2)
            
            # Create a filled rectangle for text background
            text_size = cv2.getTextSize(label, cv2.FONT_HERSHEY_SIMPLEX, 0.5, 2)[0]
            cv2.rectangle(img_with_boxes, (x1, y1 - text_size[1] - 10), (x1 + text_size[0], y1), color, -1)
            
            # Add label with contrasting text color
            # Choose black or white text based on background brightness
            brightness = (color[0] + color[1] + color[2]) / 3
            text_color = (0, 0, 0) if brightness > 127 else (255, 255, 255)
            cv2.putText(img_with_boxes, label, (x1, y1 - 5), cv2.FONT_HERSHEY_SIMPLEX, 0.5, text_color, 2)
            
            # Store detection for table with color information
            detections.append({
                "category": cls_name,
                "confidence": conf,
                "position": (x1, y1, x2, y2),
                "color": color
            })
    

        # Extract GD&T and Datum OCR information
        gdnt_info = date_extractor.read_rois(img_array, [4], boxes, 0)
        if gdnt_info is not None:
            gdnt_data.append(gdnt_info)

        surface_info = date_extractor.read_rois(img_array, [3,6,8,9,10,11], boxes, 0)
        if surface_info is not None:
            surface_data.append(surface_info)

    # If we have detected tables but no extracted images, handle that case
    if len(table_data) > 0 and len(table_images) == 0:
        table_images = [Image.fromarray(np.zeros((100, 100, 3), dtype=np.uint8))]
    
    # Return the detection result image, any extracted table image, and the JSON data
    return (
        Image.fromarray(img_with_boxes),  # Main detection image
        table_images[0] if table_images else None,  # First table image or None
        table_data[0] if len(table_data) == 1 else table_data,  # JSON data for gr.JSON
        gdnt_data[0] if len(gdnt_data) == 1 else table_data,  # JSON data for gr.JSON
        surface_data[0] if len(surface_data) == 1 else table_data  # JSON data for gr.JSON
    )

# Create Gradio interface
with gr.Blocks(title="CAD 2d Drawing Data Extraction") as app:
    gr.Markdown("# CAD 2d Drawing Data Extraction")
    gr.Markdown("Upload an image to detect objects. Tables will be automatically extracted.")
    
    with gr.Row():
        with gr.Column(scale=2):
            input_image = gr.Image(type="pil", label="Input Image")
            gr.Markdown("## Extracted Table Region")
            table_image = gr.Image()
            gr.Markdown("### Extracted GD&T Data")
            gdnt_json = gr.JSON(open=True)
            gr.Markdown("### Extracted Surface Data")
            surface_json = gr.JSON(open=True)
        with gr.Column(scale=3):
            submit_btn = gr.Button("Detect Objects", variant="primary")
            gr.Markdown("## Detection Results")
            output_image = gr.Image()
            gr.Markdown("### Extracted Table Data")
            table_json = gr.JSON(open=True)

    submit_btn.click(
        fn=process_image,
        inputs=[input_image],
        outputs=[output_image, table_image, table_json, gdnt_json, surface_json]
    )

# Launch the app
if __name__ == "__main__":
    app.launch()