A pipeline that automatically generate a summarizing clip from a video based on text description of the video

Usage

Specify necessary directories Run bash run.sh

Code Structure

1. shot_detection: segment video into shots
2. text2frame: module that does description to frames matching => find the frames that matches the best with regard to the description of the video

Attention:

Download the necessary trained model for image captioning and put to <text2frame/caption/checkpoints>