Awesome

StoryGraphs

This is a Matlab implementation of the paper:

StoryGraphs: Visualizing Character Interactions as a Timeline
Makarand Tapaswi, Martin Bäuml, and Rainer Stiefelhagen
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014
Project page | Paper download | ShotThreading & SceneDetection code

This project was inspired by the famous webcomic XKCD, in particular the narrative charts.

Tested on

Ubuntu 12.04, Ubuntu 14.04, with Matlab versions: R2013b and R2014a.
I haven't tried with the new graphics engine on Matlab R2014b and <code>draw_storygraphs.m</code> may need some changes.

First initialization

The <code>first_init.m</code> script will be called on running <code>startup.m</code> the first time. This will ask you to install three external toolboxes. Please follow the instructions.

Example usage

<code>VS = BBT(1, 1);</code>
<code>SG = storygraph(VS);</code>

Currently supported with ground-truth face identification for two TV series

• <em>The Big Bang Theory</em> (BBT) season 1, episodes 1 -- 6
• <em>Buffy the Vampire Slayer</em> (BUFFY) season 5, episodes 1 -- 6

They can be automatically generated by calling the script <code>cameraready_draw_sg.m</code>. Line 70 <code>runexp = 1;</code> controls which experiment is performed.

External toolboxes

• jsonlab: Matlab JSON interface
• export_fig: (Optional) Save generated StoryGraphs
• xticklabel_rotate: Rotate figure XTick labels (Matlab R2014b onwards has inbuilt support!)

Main functions

• storygraph.m Generates the StoryGraph for video
• storygraph_objective.m The main objective function with gradient computation
• storygraph_optimization.m Optimization wrapper for Matlab's <code>fmincon</code>
• draw_storygraph.m StoryGraph drawing function (run directly to try with random data)
• co_occurrence_scenes.m Character occurrence matrix for each scene of the video

Changelog

• 03-12-2014: v0.1: First working implementation for BBT, BUFFY