I learned from https://thescriptlab.com/screenwriting/structure/the-scene/16-types-of-scenes/ the types of scenes that a video has.
What it does
The implementation is for detecting the "Setting - Where are we?" and "Exposition – Necessary information. Quick and Clever.". Obviously there is a huge part in scene detection change as well.
How I built it
Using Python 3.6, tensorflow, Places365 dataset, opencv, etc.
What I learned
To integrate a vast diversity of python libraries.
What's next for video-scene-inventory
Implement even more detections: burned subtitles ocr for old videos, action description, etc.