Text2Image Search

This is a text2image search application built for the purpose of completing this task.

I have chosen to work with the Advertisement Image Dataset provided in the description.

In this readme you will find instructions on how to install and run the application, while the dataset exploration and approach & evaluation description documents are in separate files.

• dataset exploration
• approach and evaluation documentation

Installation and setup

Step 1

1. Set up an environment where the requirements are satisfied. Make sure to install qdrant and docker as well.

pip install -r requirements.txt

2. Install the dataset on your local machine. This is made easy with the bash script which downloads and unzips the data.

bash install_data.bash

Step 2

Now that the environment and dataset are prepared, we can run the application. Since this is running the qdrant docker command, make sure you run it as admin.

bash run_app.bash

NB This bash script abstracts two commands that would continuously run in separate terminal windows. If your python environment is located at a different location than the one in this bash script, please change it so it runs as expected.

Alternatively, run these 2 commands in 2 terminal windows from the directory of the repo:

docker run -p 6333:6333 -p 6334:6334 -v $(pwd)/qdrant_storage:/qdrant/storage:z qdrant/qdrant

streamlit run image_search_app.py

Step 3

The app will run but we need to populate the database with our encodings in order to search. This is done by simply running the encode_images.py script, and it only needs to run once.

python encode_images.py

After the script finishes running, you will be able to visit the search GUI and query the database!