import numpy as np
from tensorflow.keras.applications import ResNet50
from tensorflow.keras.preprocessing import image
from tensorflow.keras.applications.resnet50 import preprocess_input
from sklearn.metrics.pairwise import cosine_similarity
import os

# Load the pre-trained ResNet50 model
model = ResNet50(weights='imagenet', include_top=False, pooling='avg')


# Function to extract feature vector from an image
def extract_features(img_path, model):
    img = image.load_img(img_path, target_size=(224, 224))
    img_data = image.img_to_array(img)
    img_data = np.expand_dims(img_data, axis=0)
    img_data = preprocess_input(img_data)
    features = model.predict(img_data)
    return features.flatten()


# Directory containing images
image_dir = './forward_facing'

# Extract features for all images
image_features = {}
for img_file in os.listdir(image_dir):
    img_path = os.path.join(image_dir, img_file)
    features = extract_features(img_path, model)
    image_features[img_file] = features

# Convert feature dictionary to list for processing
feature_list = list(image_features.values())
file_list = list(image_features.keys())

# Calculate similarities
num_images = len(file_list)
similarity_matrix = np.zeros((num_images, num_images))

for i in range(num_images):
    for j in range(i, num_images):
        if i != j:
            similarity = cosine_similarity(
                [feature_list[i]],
                [feature_list[j]]
            )[0][0]
            similarity_matrix[i][j] = similarity
            similarity_matrix[j][i] = similarity

# Identify and remove duplicates
threshold = 0.9  # Similarity threshold for duplicates
duplicates = set()
for i in range(num_images):
    for j in range(i + 1, num_images):
        if similarity_matrix[i][j] > threshold:
            duplicates.add(file_list[j])

# Remove duplicates
# for duplicate in duplicates:
#     os.remove(os.path.join(image_dir, duplicate))
print("Duplicate Images No => ", len(duplicates))

# print(f"Removed {len(duplicates)} duplicate images.")