Professional Documents
Culture Documents
Reality
Blaine Bell
Steven Feiner
Presented By
Team Avatar
Damien Rompapas
Tobias Hllerer
Overview
Background
Summary
View Management Explained
Object Properties and Constraints
View Management Approach
My opinion
Q&A
Overview
Background
Summary
View Management Explained
Object Properties and Constraints
View Management Approach
My opinion
Q&A
View Management?
By View Management we mean the following:
Adjusting objects in our field of view to get a
better understanding of our intended target or
information in relation to the current scene
View Management?
We often make view management tasks daily, usually
Subconsciously
View Management
System aims to automatically perform such tasks to
maintain visual constraints on the projections of labels
and objects on the viewing plane so that they
- are near relative objects
- dont occlude each other.
- satisfy applied constraints
(described later)
Viewing plane?
Overview
Background
Summary
View Management Explained
Object Properties and Constraints
View Management Approach
My opinion
Q&A
Visibility
Whether an object can/cannot occlude other
defined objects on the viewing plane
Position
Min/Max distance from
Other Objects
Point/Area/Volume
Screen-Stabilized
Users Pose
World-Stabilized
World Pose
Size/Transparency
Range of possible sizes
- Preference towards the high end of range
Range of object transparency values
- Can be modified to minimize
Priority
Determine which order objects are occluded.
=> Useful for satisfying other constraints
Overview
Background
Summary
View Management Explained
Object Properties and Constraints
View Management Approach
My opinion
Q&A
View-Plane Representation
/Visible Surface Determination
Each 3D Object: Upright rectangular extent of its projection on the
view plane. (largest empty-space rectangles),
Sorted using BSP Tree for visibility representation.
Suppose that:
axis-aligned rectangular extents of the
projections of all scene objects
have been added to the representation,
and that we are given the extent A
of some new objects projection.
BSP Tree
Best known in the video game Doom
Each node stores a Hyperplane, dividing the space into two halves
Temporal Continuity
Each frame is computed independently.
Frames computed independently result in
discontinuous changes in:
- Size
- Position
This can result in jumpy and
confusing transitions
between states.
Temporal Continuity
Steps to solve using previous frame
- State Hysteresis
- Positional Stability
- Interpolation
State Hysteresis
There are several situations in which objects may change state,
resulting in a discrete visual change,
such as from an internal label to external one,
or from being displayed to not being displayed.
Positional Stability
To keep the labels clear,
They are placed without opportunity of jitter.
Two situations:
Best Possible Layout independent of
the previous layout.
Closest t Possible layout to previous
layout.
Interpolation
To minimize the effect of discontinuous jumps
Interpolate the values of all constraints between
frames
Implementation
Hardware
CPU:
1.4 GHz Pentium 4
RAM:
512 MB Ram
Graphics: SONICBlue FireGL 2
Software
OS: Windows 2K
View Management: Java
Rendering: Java 3D
View Management/Rendering
Seperated.
Overview
Background
Summary
View Management Explained
Object Properties and Constraints
View Management Approach
My opinion
Q&A
Fantastic
Very General Problem
The paper goes in-depth on
expressing this with previous work.
Very applicable
(Surprised current systems dont
implement)
Can improve on bad UI
Representation: Second Slide.
Great
Uses the world in miniature (WIM)
to assist in orientation and
labelling environment
Labels Jump between real world and
world in miniature
Great
Provides relation to research and human
habits to generate overview, easily able to
understand context of the problem.
My favorite paragraph in the paper.
In the real world, we make simple view
management decisions routinely, and often
subconsciously. For example, we push aside a
centerpiece to enable eye contact with others
at dinner, or we hold a city guidebook so that
we can read a relevant section, while
simultaneously viewing an historic building
that it describes as we pass by on a bus
Great
Detailed Description of the method
involved, extends on research from
2D environments
Open to re-implementation
(Source code not available, explained
later)
Good
Paper does show one comparison
between their labelling and other
typical labelling methods.
(Left to right: Nave,Suppression, View Management)
Good
Wide Variety of references, Some of
them are good papers!
B. Bell and S. Feiner. Dynamic space management for
user interfaces. In Proc. ACM UIST 2000
Strange
Lacks teaser image description,
Instead referred in Introduction
through repeat of images.
Most papers in the same year journal
(I checked about 30) didnt have a
teaser image, very strange
Doubtful
Lack of Usability Studies
They talk about this in future work
Doubtful
Some parts of the paper were hard
to understand
Far too wordy, slightly overexplained.
Requires background knowledge
Had to read some of the referenced
papers before beginning to understand
Doubtful
The video shown was recorded
using a HMD on a tripod.
Prevent rapid shaking
How Stable the positional stability is
Bad
They very briefly talked about, but didnt
show the GUI used for adding/creating
annotations.
(As far as the paper is aware)
We dont know how difficult it is to add
these annotations, it could all be textbased?
If its complex, it could make the system
useless for most use cases.
William Chiong and Yoav Hirsch developed the
software infrastructure that we used to create a GUI for
specifying user parameters.
Sad
This system is so good!
Why arent we using it?
Possibilities
Computation Power
Related Observation
The Labelling system generated using
this view-management system tends
to favor the outside edges of the
screen, something that many user
interfaces already do.
Why is it related?
If you added many more external
labels to the Journalism tag,
Wouldnt we end up with
a cluttered and confusing labelled
environment, akin to my second slide?
The problem may be unavoidable.
Final Opinion
What makes a system good and easy to use can
depend entirely on how information is
represented on the screen as well as the quality,
too much can be intimidating and confusing, too
little is a waste of space.
This paper opens up a new automated method
to represent such information .
Fantastic
Very general problem
Great
Detailed Description of the method involved
Good
Wide Variety of references
Strange
Lacks teaser image description
Doubtful
Lack of Usability Studies
Paper hard to understand
Bad
GUI not explained in detail
Sad
We dont use the system.