🖼️ CLIP-Powered Meme Search

A Python tool to search your meme (or image) collection using text queries!
Powered by OpenAI’s CLIP model, it finds and ranks images that match your description—even for fuzzy, creative prompts.

---

✨ Features

• ✅ Search images with natural language queries
• 🤖 Uses OpenAI’s CLIP (ViT-B/32) model for zero-shot vision-language matching
• ⚡️ Caches extracted image features for instant future searches
• 🗂️ Supports multiple image formats: .jpg, .png, .gif, .bmp
• 🖼️ Displays top matching images in notebook/console

---

📦 Tech Stack

• Python 3.x
• PyTorch
• Transformers (Hugging Face)
• NumPy
• Pillow (PIL)
• scikit-learn
• pickle
• glob

---

🚀 Getting Started

🛠️ Prerequisites

Install required packages:

pip install torch torchvision
pip install transformers
pip install numpy pillow scikit-learn

---

📁 Folder Structure

your_project/
│
├── content/
│   └── memes/
│       ├── meme1.jpg
│       ├── meme2.png
│       └── ...
├── meme_features.pkl        (auto-generated)
└── your_script.py

📌 All your images should be in content/memes/.

---

🧠 How It Works

1️⃣ Loads or extracts CLIP features for all images in your content/memes/ folder
2️⃣ Caches those features in meme_features.pkl for instant future runs
3️⃣ Processes your text query into CLIP embedding
4️⃣ Computes cosine similarity between query and images
5️⃣ Returns and displays the top matching images

---

⚙️ Example Usage

from your_module_name import search_images

query = "funny cat with sunglasses"
result_path = search_images(query, top_k=5)
print("Best match:", result_path)

---

✅ Caching

• First run: Extracts and caches features
• Next runs: Loads cached features instantly

If you add new images, simply delete meme_features.pkl to regenerate features.

---

🎥 Demo

✅ See it in action here:

👉 Watch Demo

---

🌐 Author

Md Shahriar Rahman Bhuiyan
🌐 LinkedIn Profile

---

💬 Feedback

If you have feedback or want to collaborate, feel free to reach out via LinkedIn!