PIIM IS A RESEARCH AND DEVELOPMENT
FACILITY AT THE NEW SCHOOL
© 2015 PARSONS JOURNAL FOR
INFORMATION MAPPING AND PARSONS
INSTITUTE FOR INFORMATION MAPPING
68 Fifth Avenue
New York, NY 10011
THE PARSONS INSTITUTE
FOR INFORMATION MAPPING
212 229 6825
piim.newschool.edu
KEYWORDS Collaboration, computer vision, cultural
analytics, economy of abundance, interactive data
visualization
PROJECT DATE 2014
URL http://misharabinovich.com/soyummy.html
ABSTRACT his paper describes an interdisciplinary
collaboration that began with creative misuse of artiicial
intelligence and computer vision algorithms originally
developed at McGill’s Centre for Intelligent Machines.
We began by analyzing image banks and video with
sotware originally designed for surveillance and robotic
anomaly detection. We started with basic visual analysis
and had only limited success with movie summarization.
Gaining Insight Into Films
Via Topic Modeling & Visualization
MISHA RABINOVICH, MFA
YOGESH GIRDHAR, PHD
Figure 1: Jay Muhlin Edit
We moved beyond misuse when the sotware actually
became useful for ilm analysis with the addition of audio
analysis, subtitle analysis, facial recognition, and topic
modeling. Using multiple types of visualizations and
a back-and-fourth worklow between people and AI
we arrived at an approach for cultural analytics that
can be used to review and develop ilm criticism. Finally,
we present ways to apply these techniques to Database
Cinema and other aspects of ilm and video creation.
INTRODUCTION
In the summer of 2013, Misha Rabinovich was an artist
in residence at McGill’s Centre for Intelligent Machines
working primarily with Yogesh Girdhar (then earning
his doctorate degree). Our goal was to stage ordering
interventions in an deluge of cultural information.
We began by creatively misusing robotic navigation
sotware which was originally developed for robotic
navigation and anomaly detection. his sotware worked
to summarize data by observing a data stream and saving
only those observations which would provide an overview
of the whole data set. It measured the amount of surprise