text-video-search

Text-video search app

Build a text-video search from scratch based on CLIP models with Vespa python API.

See Animation

Create the application in a Jupyter Notebook

Create, deploy, feed and query the application from a single notebook with Pyvespa.

Demo the search app

Run a local demo of the text-video search app built here.

Set the following environment variables required by the app:

export VESPA_ENDPOINT = <your-vespa-endpoint>
export VESPA_CERT_PATH= <your-vespa-certificate-path> # if using Vespa Cloud
export VIDEO_DIR=<video-folder>

Run the app:

streamlit run src/python/app.py