Build a local image search tool

Here’s a complete guide to building a local image search tool using Python. The tool will allow users to search for images in a directory using image similarity (via feature extraction) or file names.

Requirements

Install the required libraries first:

bash
CopyEdit
pip install opencv-python scikit-learn numpy pillow matplotlib

Python Script: Local Image Search Tool

python
CopyEdit
import os
import cv2
import numpy as np
from sklearn.metrics.pairwise import cosine_similarity
from PIL import Image
import matplotlib.pyplot as plt

# Directory containing your local images
IMAGE_DIR = "images"

# Supported image formats
SUPPORTED_FORMATS = (".jpg", ".jpeg", ".png", ".bmp", ".tiff")

def load_images(image_dir):
    images = {}
    for filename in os.listdir(image_dir):
        if filename.lower().endswith(SUPPORTED_FORMATS):
            path = os.path.join(image_dir, filename)
            image = cv2.imread(path)
            images[filename] = image
    return images

def extract_features(image, size=(100, 100)):
    image = cv2.resize(image, size)
    image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    return image.flatten()

def build_feature_database(images):
    feature_db = {}
    for name, img in images.items():
        feature_db[name] = extract_features(img)
    return feature_db

def search_similar_images(query_image_path, feature_db, top_k=5):
    query_image = cv2.imread(query_image_path)
    query_features = extract_features(query_image)
    similarities = {}

    for name, features in feature_db.items():
        sim = cosine_similarity([query_features], [features])[0][0]
        similarities[name] = sim

    sorted_results = sorted(similarities.items(), key=lambda x: x[1], reverse=True)
    return sorted_results[:top_k]

def show_images(result_list, images):
    plt.figure(figsize=(15, 5))
    for i, (name, score) in enumerate(result_list):
        img = cv2.cvtColor(images[name], cv2.COLOR_BGR2RGB)
        plt.subplot(1, len(result_list), i + 1)
        plt.imshow(img)
        plt.title(f"{name}nScore: {score:.2f}")
        plt.axis("off")
    plt.show()

def main():
    print("Loading image database...")
    images = load_images(IMAGE_DIR)
    feature_db = build_feature_database(images)

    query_path = input("Enter path to query image: ")
    if not os.path.exists(query_path):
        print("Image not found.")
        return

    results = search_similar_images(query_path, feature_db)
    print("nTop Matches:")
    for name, score in results:
        print(f"{name} (Similarity: {score:.2f})")

    show_images(results, images)

if __name__ == "__main__":
    main()

Features

• Image content-based search using grayscale histogram vectors.

• Top-k similarity ranking with cosine similarity.

• Visual preview using matplotlib.

• Easily expandable with more advanced models like ResNet for deeper features.

Optional Enhancements

1. Add filename keyword search:

   python
   CopyEdit
   def search_by_filename(keyword, image_db):
       return {name: img for name, img in image_db.items() if keyword.lower() in name.lower()}
   

2. Use deep learning features (e.g., from VGG16 or ResNet with torchvision or keras.applications) for higher accuracy in visual similarity.

3. Web Interface using Flask or Streamlit.

Let me know if you’d like a Streamlit or Flask version to turn it into a GUI/web app.