Introduction
This code implements a versatile image search engine leveraging the CLIP model and FAISS, capable of processing both text-to-image and image-to-image queries.
Image to image
Text to image
Install Environment
pip install -r requirements.txt
Preprocess
python clip_image_search/extract_embeddings.py --img_dir your_dataset_dir --save_path results/embeddings.pkl
python clip_image_search/build_index.py --embeddings_path results/embeddings.pkl --save_path results/index.faiss
Demo
streamlit run app.py