app.py

import gradio as gr
from ultralytics import YOLO
import cv2
import numpy as np
from PIL import Image, ImageDraw, ImageFont
import sqlite3
import base64
from io import BytesIO
import tempfile
import pandas as pd

# Load YOLOv8 model
model = YOLO("best.pt")

def predict_image(input_image, name, patient_id):
    if input_image is None:
        return None, "Please Input The Image"

    # Convert Gradio input image (PIL Image) to numpy array
    image_np = np.array(input_image)

    # Ensure the image is in the correct format
    if len(image_np.shape) == 2:  # grayscale to RGB
        image_np = cv2.cvtColor(image_np, cv2.COLOR_GRAY2RGB)
    elif image_np.shape[2] == 4:  # RGBA to RGB
        image_np = cv2.cvtColor(image_np, cv2.COLOR_RGBA2RGB)

    # Perform prediction
    results = model(image_np)

    # Draw bounding boxes on the image
    image_with_boxes = image_np.copy()
    raw_predictions = []

    if results[0].boxes:
        # Sort the results by confidence and take the highest confidence one
        highest_confidence_result = max(results[0].boxes, key=lambda x: x.conf.item())

        # Determine the label based on the class index
        class_index = highest_confidence_result.cls.item()
        if class_index == 0:
            label = "Immature"
            color = (0, 255, 255)  # Yellow for Immature
        elif class_index == 1:
            label = "Mature"
            color = (255, 0, 0)  # Red for Mature
        else:
            label = "Normal"
            color = (0, 255, 0)  # Green for Normal

        confidence = highest_confidence_result.conf.item()
        xmin, ymin, xmax, ymax = map(int, highest_confidence_result.xyxy[0])
        
        # Draw the bounding box
        cv2.rectangle(image_with_boxes, (xmin, ymin), (xmax, ymax), color, 2)
        
        # Enlarge font scale and thickness
        font_scale = 1.0
        thickness = 2
        
        # Calculate label background size
        (text_width, text_height), baseline = cv2.getTextSize(f'{label} {confidence:.2f}', cv2.FONT_HERSHEY_SIMPLEX, font_scale, thickness)
        cv2.rectangle(image_with_boxes, (xmin, ymin - text_height - baseline), (xmin + text_width, ymin), (0, 0, 0), cv2.FILLED)
        
        # Put the label text with black background
        cv2.putText(image_with_boxes, f'{label} {confidence:.2f}', (xmin, ymin - 10), cv2.FONT_HERSHEY_SIMPLEX, font_scale, (255, 255, 255), thickness)
        
        raw_predictions.append(f"Label: {label}, Confidence: {confidence:.2f}, Box: [{xmin}, {ymin}, {xmax}, {ymax}]")

    raw_predictions_str = "\n".join(raw_predictions)
    
    # Convert to PIL image for further processing
    pil_image_with_boxes = Image.fromarray(image_with_boxes)

    # Add text and watermark
    pil_image_with_boxes = add_text_and_watermark(pil_image_with_boxes, name, patient_id, label)
    
    return pil_image_with_boxes, raw_predictions_str

# Function to add watermark
def add_watermark(image):
    try:
        logo = Image.open('image-logo.png').convert("RGBA")
        image = image.convert("RGBA")
        
        # Resize logo
        basewidth = 100
        wpercent = (basewidth / float(logo.size[0]))
        hsize = int((float(wpercent) * logo.size[1]))
        logo = logo.resize((basewidth, hsize), Image.LANCZOS)
        
        # Position logo
        position = (image.width - logo.width - 10, image.height - logo.height - 10)
        
        # Composite image
        transparent = Image.new('RGBA', (image.width, image.height), (0, 0, 0, 0))
        transparent.paste(image, (0, 0))
        transparent.paste(logo, position, mask=logo)
        
        return transparent.convert("RGB")
    except Exception as e:
        print(f"Error adding watermark: {e}")
        return image

# Function to add text and watermark
def add_text_and_watermark(image, name, patient_id, label):
    draw = ImageDraw.Draw(image)
    
    # Load a larger font (adjust the size as needed)
    font_size = 48  # Example font size
    try:
        font = ImageFont.truetype("font.ttf", size=font_size)
    except IOError:
        font = ImageFont.load_default()
        print("Error: cannot open resource, using default font.")
    
    text = f"Name: {name}, ID: {patient_id}, Result: {label}"
    
    # Calculate text bounding box
    text_bbox = draw.textbbox((0, 0), text, font=font)
    text_width, text_height = text_bbox[2] - text_bbox[0], text_bbox[3] - text_bbox[1]
    text_x = 20
    text_y = 40
    padding = 10

    # Draw a filled rectangle for the background
    draw.rectangle(
        [text_x - padding, text_y - padding, text_x + text_width + padding, text_y + text_height + padding],
        fill="black"
    )
    
    # Draw text on top of the rectangle
    draw.text((text_x, text_y), text, fill=(255, 255, 255, 255), font=font)

    # Add watermark to the image
    image_with_watermark = add_watermark(image)
    
    return image_with_watermark

# Function to initialize the database
def init_db():
    conn = sqlite3.connect('results.db')
    c = conn.cursor()
    c.execute('''CREATE TABLE IF NOT EXISTS results
                 (id INTEGER PRIMARY KEY, name TEXT, patient_id TEXT, input_image BLOB, predicted_image BLOB, result TEXT)''')
    conn.commit()
    conn.close()

# Function to submit result to the database
def submit_result(name, patient_id, input_image, predicted_image, result):
    conn = sqlite3.connect('results.db')
    c = conn.cursor()
    
    input_image_np = np.array(input_image)
    _, input_buffer = cv2.imencode('.png', cv2.cvtColor(input_image_np, cv2.COLOR_RGB2BGR))
    input_image_bytes = input_buffer.tobytes()
    
    predicted_image_np = np.array(predicted_image)
    predicted_image_rgb = cv2.cvtColor(predicted_image_np, cv2.COLOR_RGB2BGR)  # Ensure correct color conversion
    _, predicted_buffer = cv2.imencode('.png', predicted_image_rgb)
    predicted_image_bytes = predicted_buffer.tobytes()
    
    c.execute("INSERT INTO results (name, patient_id, input_image, predicted_image, result) VALUES (?, ?, ?, ?, ?)", 
              (name, patient_id, input_image_bytes, predicted_image_bytes, result))
    conn.commit()
    conn.close()
    return "Result submitted to database."

# Function to load and view database
def view_database():
    conn = sqlite3.connect('results.db')
    c = conn.cursor()
    c.execute("SELECT name, patient_id, input_image, predicted_image FROM results")
    rows = c.fetchall()
    conn.close()
    
    # Convert to pandas DataFrame for better display in Gradio
    df = pd.DataFrame(rows, columns=["Name", "Patient ID", "Input Image", "Predicted Image"])
    
    return df

# Function to download database or image
def download_file(choice):
    if choice == "Database (.db)":
        # Provide the path to the database file
        return 'results.db'
    else:
        conn = sqlite3.connect('results.db')
        c = conn.cursor()
        c.execute("SELECT predicted_image FROM results ORDER BY id DESC LIMIT 1")
        row = c.fetchone()
        conn.close()
        if row:
            image_bytes = row[0]
            with tempfile.NamedTemporaryFile(delete=False, suffix='.png') as temp_file:
                temp_file.write(image_bytes)
                temp_file.flush()  # Ensure all data is written before closing
                return temp_file.name
        else:
            raise FileNotFoundError("No images found in the database.")

# Initialize the database
init_db()

# Gradio Interface
def interface(name, patient_id, input_image):
    if input_image is None:
        return "Please upload an image."

    output_image, raw_result = predict_image(input_image, name, patient_id)
    submit_status = submit_result(name, patient_id, input_image, output_image, raw_result)

    return output_image, raw_result, submit_status

# View Database Function
def view_db_interface():
    df = view_database()
    return df

# Download Function
def download_interface(choice):
    try:
        file_path = download_file(choice)
        with open(file_path, "rb") as file:
            return file.read(), file_path.split('/')[-1]
    except FileNotFoundError as e:
        return str(e), None

# Gradio Blocks
with gr.Blocks() as demo:
    with gr.Column():
        gr.Markdown("# Cataract Detection System")
        gr.Markdown("Upload an image to detect cataract and add patient details.")
        gr.Image("PR_curve.png", label="Model PR Curve")
        gr.Markdown("This application uses YOLOv8 with mAP=0.981")
    
    with gr.Column():
        name = gr.Textbox(label="Name")
        patient_id = gr.Textbox(label="Patient ID")
        input_image = gr.Image(type="pil", label="Upload an Image", image_mode="RGB")

    with gr.Column():
        submit_btn = gr.Button("Submit")
        output_image = gr.Image(type="pil", label="Predicted Image")

    with gr.Row():
        raw_result = gr.Textbox(label="Raw Result", lines=5)
        submit_status = gr.Textbox(label="Submission Status")

    submit_btn.click(fn=interface, inputs=[name, patient_id, input_image], outputs=[output_image, raw_result, submit_status])

    with gr.Column():
        view_db_btn = gr.Button("View Database")
        db_output = gr.Dataframe(label="Database Records")

    view_db_btn.click(fn=view_db_interface, inputs=[], outputs=[db_output])

    with gr.Column():
        download_choice = gr.Radio(["Database (.db)", "Predicted Image (.png)"], label="Choose the file to download:")
        download_btn = gr.Button("Download")
        download_output = gr.File(label="Download File")

    download_btn.click(fn=download_interface, inputs=[download_choice], outputs=[download_output])

# Launch the Gradio app
demo.launch()