import numpy as np
import gradio as gr
from PIL import Image
from skimage.metrics import structural_similarity as ssim
import cv2
import pytesseract
from difflib import SequenceMatcher

# Function to calculate SSIM between two images
def calculate_ssim(img1, img2):
    img1_gray = cv2.cvtColor(img1, cv2.COLOR_RGB2GRAY)
    img2_gray = cv2.cvtColor(img2, cv2.COLOR_RGB2GRAY)
    return ssim(img1_gray, img2_gray)

# Function to compare trademarks based on text similarity
def compare_text(trademark1, trademark2):
    text1 = pytesseract.image_to_string(trademark1).lower()
    text2 = pytesseract.image_to_string(trademark2).lower()
    similarity_ratio = SequenceMatcher(None, text1, text2).ratio()
    return similarity_ratio

# Function to compare trademarks based on color similarity
def compare_colors(trademark1, trademark2):
    trademark1 = trademark1.convert("RGB")
    trademark2 = trademark2.convert("RGB")
    colors1 = trademark1.getcolors(trademark1.size[0] * trademark1.size[1])
    colors2 = trademark2.getcolors(trademark2.size[0] * trademark2.size[1])
    color_vector1 = np.zeros(3)
    color_vector2 = np.zeros(3)
    for count, color in colors1:
        color_vector1 += np.array(color) * count
    for count, color in colors2:
        color_vector2 += np.array(color) * count
    color_vector1 /= trademark1.size[0] * trademark1.size[1]
    color_vector2 /= trademark2.size[0] * trademark2.size[1]
    color_similarity = 1 - np.linalg.norm(color_vector1 - color_vector2)
    return color_similarity

# Function to compare trademarks based on multiple aspects
def compare_trademarks(trademark1, trademark2):
    # Resize images to a consistent size
    size = (300, 300)
    trademark1_resized = trademark1.resize(size)
    trademark2_resized = trademark2.resize(size)

    ssim_score = calculate_ssim(np.array(trademark1_resized), np.array(trademark2_resized))
    text_similarity = compare_text(trademark1, trademark2)
    color_similarity = compare_colors(trademark1, trademark2)

    # Adjust weights based on the importance of each aspect
    ssim_weight = 0.6
    text_weight = 0.2
    color_weight = 0.2

    overall_similarity = (ssim_weight * ssim_score) + (text_weight * text_similarity) + (color_weight * color_similarity)
    return overall_similarity

# Function to perform trademark conflict prevention
def prevent_trademark_conflict(trademark1, trademark2):
    similarity_score = compare_trademarks(trademark1, trademark2)
    result = f"Trademark Similarity Score: {similarity_score:.4f}"
    return result

iface = gr.Interface(
    fn=prevent_trademark_conflict,
    inputs=[
        gr.inputs.Image(type="pil", label="Trademark Image 1"),
        gr.inputs.Image(type="pil", label="Trademark Image 2")
    ],
    outputs="text",
    title="Trademark Conflict Prevention",
  
    description="Upload two trademark images to prevent conflict."
)

iface.launch()