A powerful visual similarity search app using the CLIP model to match uploaded query images against a gallery — all within an interactive Streamlit UI.
✅ Built with:
CLIP,Transformers,Streamlit,PIL,PyTorch,scikit-learn
- 🖼️ Upload a query image
- 🗂️ Upload a set of gallery images
- 🔍 Get top 3 visually similar matches using CLIP embeddings
- 📊 Similarity scores shown using cosine similarity
- 🎨 Clean, minimal Streamlit UI
- Python 3.8 or higher
# 1. Clone this repo
git clone https://github.com/rakshath66/clipfindr.git
cd clipfindr
# 2. (Optional) Create a virtual environment
python -m venv venv
source venv/bin/activate # or venv\Scripts\activate on Windows
# 3. Install dependencies
pip install -r requirements.txt
# 4. Run the app
streamlit run app.pyclipfindr/
├── app.py # Streamlit app with CLIP visual search
├── gallery/ # Uploaded gallery images (auto-created)
├── requirements.txt # Python dependencies
├── README.md # This file
- 🧠 CLIP from OpenAI
- 🤗 Hugging Face Transformers
- 🧮 PyTorch + TorchVision
- 📊 Cosine Similarity via scikit-learn
- 🎨 PIL for image handling
- 🖼️ Streamlit for the frontend
- Product image deduplication
- Visual search for screenshots
- Reverse lookup from dataset images
- Similar fashion or object search
CLIP is loaded directly via Hugging Face — no tokens required.
Optional: To avoid rate limits, run
huggingface-cli loginif you have an account.
- Fork the repo
- Create a branch:
git checkout -b my-feature - Make changes and commit:
git commit -m "Add: new feature" - Push:
git push origin my-feature - Open a pull request ✅
Clean, modular contributions welcome!
MIT License © Rakshath U Shetty
- CLIP-based similarity matching
- Top 3 results with similarity scores
- Clean Streamlit UI
- ✅ Save image metadata
- 🧠 Add text + image matching
- 🖼️ Visual heatmap of similarity
- 🌐 Optional: add BLIP-based captioning