CLIP Tasks

A set of tasks CLIP can help perform without training, combining text and imagery.

A mashup of a few CLIP repos and my own code, based on:

• https://github.com/woctezuma/heroku-clip
• https://github.com/haofanwang/natural-language-joint-query-search

Install

Clone the repo! Then,

pip install -r requirements.txt

Image Labeling

First, make an outputs directory under clip-tasks. Then, In tasks/infer_image_label. py, edit the list of images and list of text queries, and then run it.

Plots should be saved in the outputs directory.

Text to Image Search

First, make an images directory under clip-tasks. Then, In tasks/text_to_image_search. py, edit the list of images and list of text queries, and then run it.

The matched image should be saved in the images directory.

Image to Image Search

Soon!

Image + Text to Image Search

Soon!

Text Video Search

streamlit run tasks/video_search_app.py --server.port 80

Image to Video Search

Soon!

Image + Text to Video Search

Soon!