Semantic Search
This repository contains a barebones implementation of a semantic search engine. The implementation is based on leveraging pre-trained embeddings from VGG16 (trained on Imagenet), and GloVe (trained on Wikipedia).
It allows you to:
Find similar images to an input image
Find similar words to an input word
Search through images using any word
Generate tags for any image
See examples of usage by following along on this notebook. Read more details about why and how you would use this in this blog post.
Setup
Clone the repository locally and create a virtual environment (conda example below):
conda create -n semantic_search python=3.5 -y
source activate semantic_search
cd semantic_search
pip install -r requirements.txt
If you intend to use text, download pre-trained G