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

Star History