Home

High Dimensional Data Visualization:

An Oculus-Rift Based Tool for 3D Visualization & Manipulation of Tabular Data

Many scientific investigations involve analyzing information in the form of a set of attributes of large numbers of particular items -- for example, people, products, cities or regions, etc. Within this context, we can consider the data as constituting a table, where a row correspond to a particular item (e.g., a particular person), and the columns represent different “fields” (or “variables”), each corresponding to an attribute of that item. For example, we might have rows representing people, and the columns representing the age, income, sex, education level, marital status, province of residence, and ethnicity of that person. In a large and important set of cases (e.g., survey data), there can be dozens to hundreds of fields (columns), and thousands to hundreds of thousands of rows. This document is focused on such cases.

Within many such contexts, there is a strong interest in interactive exploration of the data -- exploration that seeks to quickly examine patterns across the items (rows) as manifested in successively varying subsets of fields. Within this context, the user will generally plot out data for the full set of items, but quickly change from one set of fields to another subset in a fluid way. Often investigators are looking for clustering, or ways in which different subsets of people differ from one another, or co-variation (variation together) of different sets of variables. For example, given age and income as fields, an investigator might form a (two-dimensional) scatterplot to investigate how income tends to vary with age. Within this representation, each person (item) is associated with a point in (X,Y) space on the scatterplot. For this example, the X location for a person is given by that person’s age, and the Y location is given by their income. Such visualizations can take advantage of the exceptionally strong visual processing abilities of the human visual cortex, which allow us to quickly spot and (via adjustment of the fields visualized or subsets of the data considered) better tease out trends and patterns. Unfortunately, traditional data visualization packages fall far short of taking full advantage of the human visual systems, and exploration capabilities. But in other cases, such exploration is greatly limited in two respects. One of these limitations concerns dimensions of the data. When these analyses are conducted by single users in one or two dimensions (e.g., considering only age, or age and income, or age and education), these needs tend to be well supported by some existing software -- e.g., the popular platform Tableau and Apache Kibana. While the human visual system is highly evolved for visualization of objects in 3D space, traditional data visualization tools have difficulty compellingly visualizing data in 3 dimensions.

A second major limitation of existing tools reflects the fact that, in many contexts, data exploration and interaction around such analysis routinely often takes place within geographically distributed teams. Unfortunately traditional tools requires that users be located at the same location, and fail to support joint viewing of manipulations of the data by multiple parties.

The project proposed here involves building a web based real-time collaborative interface to allow multiple users for interactively creating, viewing, and altering 3D versions of tabular data using the Oculus Rift virtual reality platform.

• Project description copied from specifications supplied by professor Nathaniel Osgood for CMPT371