AUTOMATIC PLAYBACK OF 3D BLOG ENTRIES BASED ON

3D VIEWPOINT SIMILARITY

Rieko Kadobayashi

National Institute of Information and Communications Technology, 3-5 Hikaridai, Seika-cho, Kyoto 619-0289 Japan

Keywords:

3D Blog, 3D viewpoint information, Community support, Blog search, TrackBack, Viewpoint similarity.

Abstract:

The 3D Blog system allows a user to interact with a 3D model, annotate any spot on the model, and publish the

annotations in a blog so that the user can easily communicate with other people and share interests, idea, and

questions about 3D models of interest to the user. The main features of the system is that each annotation is

saved as a blog entry with 3D viewpoint information, i.e., the user’s viewpoint and view direction as well as 3D

information on annotated points. With this 3D viewpoint information, readers of a 3D Blog can reconstruct the

exact same 3D scene the blogger saw when reading the blog entry, and, therefore, the readers may understand

it better. The system also uses the 3D viewpoint information to detect similar entries and sends trackback

links automatically to enable users to easily ﬁnd people who may have similar interests. Users can easily and

intuitively compare the similar entries using the “playback” function which renders the 3D scene in the similar

entries. The 3D Blog system is expected to promote emergence of communities of people who share interests

and support their communication.

1 INTRODUCTION

The number of 3D Computer Graphics (CG) models

available on the Internet is rapidly increasing due both

to strong demand for 3D models in such domains as

online catalogues and cultural heritage sites and to the

advancement of digitizing and CG technology. For

example, in the cultural heritage sector, there is an

increasing demand to be able to record and visual-

ize cultural heritage items in 3D CG models because

these models have many advantages over traditional

recording and presentation media such as 2D draw-

ings. In online catalogues, 3D models can provide

consumers with more detailed information and inter-

activity than photographs can. This means that users

can change points of view as freely as if they were

looking at an object in a real space, and change the

texture or color of a 3D model as many times as they

want to test how its appearance changes, which may

actually be more difﬁcult in real space.

The number of blog users (bloggers) is also

rapidly increasing. Blogs are now a very important

and widespread medium for people to express and ex-

change ideas. One of the key features of a blog is that

it allows readers to comment on blog entries, meaning

that the owner of a blog site and its readers can eas-

ily communicate with each other. This bi-directional

communication is facilitated by “TrackBack” mecha-

nism which some blog systems use.

However, the use of 3D models is still restricted.

Many 3D model applications provide users with some

interactivity but do not provide users with a commu-

nication function for using the models. Assume that

you are browsing an online catalogue and you like the

design of a particular object. You may need to write

a long sentence that explains why you think the de-

sign is excellent or take a snapshot of your computer

display of the 3D model of the object if you want to

recommend it to your friend. If the online catalogue

supports communication using 3D models, the task is

much easier, and you may only need to send a link to

a 3D view of the object in a particular orientation and

a comment to your friend.

Another example in the cultural heritage sector

arises when people visit a museum. Visitors can usu-

ally share the experience with their family, friends,

Kadobayashi R. (2007).

AUTOMATIC PLAYBACK OF 3D BLOG ENTRIES BASED ON 3D VIEWPOINT SIMILARITY.

In Proceedings of the Third International Conference on Web Information Systems and Technologies - Web Interfaces and Applications, pages 18-25

DOI: 10.5220/0001264100180025

 SciTePress

or schoolmates, watching how they interact with ex-

hibits. This observational or social learning is thought

to be very important (Falk and Dierking, 2000). Shar-

ing knowledge, interests, impressions, and experi-

ences can deepen each person’s understanding and ap-

preciation of the exhibits. In addition, there are a lot

of hands-on exhibits in real museums that enable vis-

itors to interact, which facilitates their understanding.

However, existing 3D applications support only

limited sharing and interactivity. This is especially

true of applications for the general public. Experts

can interact with 3D models using proprietary or ex-

pensive commercial software, but the general public

is allowed only to view them passively, e.g. to see still

images or animation in most cases. The experts may

convey their interpretation of cultural heritage via 3D

visualization, but the general public does not have any

such opportunity.

Of course, there are 3D applications targeting the

general public such as virtual 3D museums. For ex-

ample virtual Leonardo provides on-line chat func-

tions so that users can communicate with each other

(Paolini et al., 1999). The main problem in vir-

tual 3D museums, from the viewpoint of communi-

cation and collaboration, is that users need to be in

a virtual museum at the same time. Practically, this

may be an overly restrictive requirement, and other

means of collaboration should be provided (Barbieri

and Paolini, 2000).

To overcome these problems, the 3D Blog system

has been proposed. This system allows a user to inter-

act with a 3D model, annotate any spot on the model,

and publish the annotations in a blog so that the user

can easily communicate with other people and share

interests, ideas, and questions about 3D models of in-

terest to the user (Kadobayashi, 2005). One of the

main features of a 3D blog is that each annotation

is saved as a blog entry with 3D information on the

user’s viewpoint. Readers of a 3D Blog can recon-

struct the exact same 3D scene the blogger saw when

reading the blog entry, perhaps helping the readers to

understand it better.

In this paper, we propose a new mechanism for

viewing 3D Blog entries that enables users to ﬁnd

similar entries easily and that intuitively calls their at-

tention to the similarities among entries by using 3D

models. It also enables them to ﬁnd people who may

have similar interests and to promote emergence of

communities of people who share interests.

The rest of this paper is organized as follows. Sec-

tion 2 introduces the 3D Blog system and describes

3D viewpoint information. In Section 3, the algo-

rithm for ﬁnding similar entries and the mechanism

that links them are described. In Section 4 the al-

gorithm for playing back similar entries is explained.

Section 5 discusses the features of the 3D Blog sys-

tem that make it a very useful communication tool.

Section 6 concludes the paper.

2 THE 3D BLOG SYSTEM

Entry

Snapshot

List of 3D

models

Figure 1: Top page of a 3D Blog. Annotations to 3D mod-

els are listed as blog entries in reverse chronological order.

Each entry has a snapshot of the 3D scene the user saw when

annotating.

2.1 Features

3D Blog enables users to interact with a 3D model,

annotate any spot on the 3D model, and publish the

annotations as a blog. It does not require any ex-

pensive proprietary software to make annotations or

read blog entries. It only requires a 3D model viewer

that runs as a browser plug-in for interacting with

3D models. The current 3D Blog system uses lily

(lily, 2005) as a blogging tool and Cortona VRML

Client (ParallelGraphics, 2006) as a viewer for the 3D

models. Lily is a Ruby implementation of blosxom

(Blosxom, 2006) written in Perl. Users need to down-

load and install the plug-in, i.e., Cortona VRML

Client, to their web browser (we are currently using

Internet Explorer only) before they start 3D blogging.

The main feature of 3D Blog is that it uses the

3D viewpoint information of the 3D scene that the

user (blogger) sees when adding an annotation. The

3D viewpoint information is composed of viewpoint

and view direction, which are expressed as 3D coor-

dinates, and a 3D vector. The 3D viewpoint infor-

mation, as well as the position of an annotated spot,

AUTOMATIC PLAYBACK OF 3D BLOG ENTRIES BASED ON 3D VIEWPOINT SIMILARITY

is automatically added to the annotation by the sys-

tem and stored in the database. It is then used to re-

construct the same 3D scene the blogger saw when a

reader reads entries. This is thought to greatly help

readers to understand the content of the entries.

The viewpoint is the point from which a user looks

at the point being annotated. The view vector is a vec-

tor from the viewpoint to the point being annotated.

The view direction is derived by normalizing the view

vector. In implementation, these items are altered by

the parameters speciﬁc to a computer graphics soft-

ware. For example, in VRML, which we used in our

system, the viewpoint and the view direction are ex-

pressed as parameters “translation” and “rotation” of

a camera.

This system is based on the assumption that im-

pressions may change if the user’s point of view

changes, even if the user looks at the same thing. The

resulting design may enable users to convey their feel-

ings more accurately if they can show the object as it

looked when they saw it.

The 3D viewpoint information is also useful for

searching (Kadobayashi and Tanaka, 2005; Fujita and

Arikawa, 2005) and in the 3D Blog system is used

to ﬁnd similar entries. When a new entry is posted,

the system detects similar entries based on a thresh-

old set by users and sends links from the new entry

to similar entries. We call this mechanism “automatic

TrackBack.” It helps users ﬁnd other users with sim-

ilar interests and create communities and facilitates

communication among users.

To summarize, by embedding 3D viewpoint infor-

mation into blog entries, the system enables users to

easily understand the idea of blog entries by playing

back a viewpoint of the annotated point, ﬁnding sim-

ilar entries, and communicating with similar users,

which facilitates forming communities.

2.2 Writing 3d Blog Entries

When a 3D Blog user logs in, the top page of the

user’s blogsite will open and show recent entries in

reverse chronological order, as in ordinary blogsites

(Figure 1). By selecting a 3D model from the list on

the left-hand side of the window and clicking on the

model’s name, the 3D model viewer will appear at the

top of the list of entries enabling the user to browse

and interact with the model in the walk-through, ﬂy-

through, or study mode and to add notes about the

model (Figure 2).

When the user clicks on a spot that he/she wants to

annotate, a small ball, i.e. a marker, appears and cov-

ers the spot in the 3D model viewer. The user types

an appropriate title and a comment in the dialogue box

3D model viewer

Text input box

Entries on

this 3D model

Figure 2: Typical page layout for interacting with and an-

notating 3D models. 3D model viewer is displayed on top

of entry list.

below the 3D model viewer and then clicks the ‘post’

button.

Information, such as the position of the annotated

area, the viewpoint of the virtual camera which is

treated as that of the user, the date and time, the snap-

shot of the 3D model with the marker, and the anno-

tation (i.e., the title and comments) are converted into

a blog entry and stored in the blog database. The an-

notation is then added to the top of the entries of the

user’s blog site as the latest entry.

Users can annotate a 3D model as often as they

desire. All annotations are automatically added to the

blog site as separate entries. The user can change 3D

models by clicking on a model’s name and continuing

to blog on a different model.

2.3 Reading 3d Blog

At ﬁrst glance, the top page of a 3D blogsite looks

similar to the top pages of an ordinary blog, but there

is a big difference. Every entry in the 3D blog in-

cludes information about the 3D viewpoint, i.e., the

viewpoint from which the 3D model was annotated

and about the spot to which the annotation was added.

The snapshot included in each entry is an an-

chor to the 3D scene and, once it is clicked by a

reader, the 3D model viewer pops up to render the 3D

scene using the 3D viewpoint information while at the

same time enabling the reader to read the entry (Fig-

ure 3). When rendering the 3D scene, the system ﬁrst

WEBIST 2007 - International Conference on Web Information Systems and Technologies

3D model viewer

Anchor to the 3D scene

Figure 3: 3D model viewer pops up when a snapshot of an

entry is clicked since a javascript is embedded as an anchor

to the 3D scene. The user can read the entry and conﬁrm

how the 3D model looks.

shows the 3D model in a default orientation and then

changes the orientation of the model so that the 3D

scene is exactly the same as the blogger saw. This an-

imation brings to the reader’s attention how the blog-

ger changed viewpoints when annotating. This is a

great help to readers in understanding the content of

entries, because it enables them to visually conﬁrm it.

Readers can add comments to an entry by click-

ing the ‘comment’ button at the bottom of the entry

to open a dialogue box. Other basic functions such

as viewing old entries and searching entries are also

available.

3 FINDING SIMILAR ENTRIES

The 3D Blog system uses the 3D viewpoint informa-

tion to ﬁnd similar entries and to render 3D scenes

and calls user attention to these similar entries by au-

tomatically setting TrackBack links. There is a blog

service that connects entries that share the same key-

words (hatena, 2007), i.e., semantic similarities. In

contrast, the 3D Blog system uses the physical per-

spective similarities.

In the current implementation, four measures are

used to detect similar entries: target, viewpoint, direc-

tion, and gaze. Target is the distance between anno-

tated points while viewpoint is the distance between

viewpoints. Similarity in direction is obtained by co-

(1) Targets are similar.

(2) Viewpoints are similar.

(3) Directions are similar.

(4) Gaze is similar.

Figure 4: Similarity measures.

sine similarity of view direction vectors. Since the

value of cosine theta is not linear, it is not reasonable

to expect users to set the threshold. As a result, the

degree of the angle between two vectors is used as

a measure and users set the threshold from 0 to 180

degrees. Gaze is a combined measure of target and

viewpoint. If target is less than or equal to the thresh-

old and viewpoint is less than or equal to the thresh-

old, the entries are similar in terms of gaze. Figure 4

depicts examples that are similar for each measure.

Usually the TrackBack link is used by a blogger

who writes an entry referring to another blogger’s en-

AUTOMATIC PLAYBACK OF 3D BLOG ENTRIES BASED ON 3D VIEWPOINT SIMILARITY

TrackBack

User A User B(1) (2)

(3)

Figure 5: Automatic TrackBack is conducted by the system

when similar entries are found. Users may easily ﬁnd other

users with similar interests.

try to explicitly notify the other blogger of the refer-

ence. In the 3D Blog system, the system also auto-

matically sends TrackBack to entries in the database

if it ﬁnds similar entries as new entries are written.

The system checks the similarities between each

new entry and the stored entries whenever a new en-

try is posted. Then it makes the TrackBack links from

the new entry to the stored entries. This means the

latest entry by any user of the 3D Blog system has no

automatic TrackBack links. The 3D Blog system in-

cludes the information about the similarity measures,

i.e., the similarity score, as well as such information

as the title and URL of the blog, which is normally in-

cluded in the TrackBack ping (SixApart, 2004). The

blogger’s own entries are excluded from the similar-

ity check since it is unremarkable that entries by the

same blogger tend to be similar.

This automatic TrackBack mechanism helps 3D

Blog users ﬁnd similar entries and others who may

have similar interests or perspectives. Users can set

the threshold according to their preference. For exam-

ple, some people may want to ﬁnd people who have

similar interests and set a low threshold for the dis-

tance between annotated spots. Others may want to

ﬁnd people who have a similar (physical) perspective

and set that distance threshold low.

3.1 Finding Similar Entries: A Scenario

Assume that user A has written an entry about a 3D

model and the entry is stored in the database (see

(1) in Figure 5). After user B writes an entry about

the same 3D model (see (2) in Figure 5), the system

Similarity measures

Target

Viewpoint

Direction

Gaze

Score

Similar

entries

target = 0.22 viewpoint = 9.1 direction = 71.92 gaze = 9.1

Figure 6: List display of similar entries. Entries are sorted

in ascending order of the score of the measure selected by

the user. In this case entries are sorted using the score of

“target” measure.

checks for similar entries in the database. In this case,

the system ﬁnds that an entry written by user A is sim-

ilar to the entry that user B has just posted because the

distance between the annotated spots of the two en-

tries is very small and is less than the threshold spec-

iﬁed by user A. It then sends TrackBack to user A’s

entry so that user A will notice user B’s entry (see (3)

in Figure 5).

4 AUTOMATIC PLAYBACK OF

3D BLOG ENTRIES

The 3D Blog system enables a user to see similar en-

tries in playback mode as well as in list mode. The

list mode, which is shown in Figure 6, looks similar

to many blog pages. The user’s entry is shown on top

of the blog page and similar entries written by other

bloggers are shown below the user’s entry. Users can

choose the sort key from the four measures. For ex-

ample, if target is chosen, the entries are sorted in

ascending order of the target score.

In the playback mode, by contrast, the user can

playback the 3D scene in each similar entry so that

she or he will notice which spots on a 3D model are

of interest to other users and which of those spots are

annotated by other users. First, the user chooses “user

WEBIST 2007 - International Conference on Web Information Systems and Technologies

Markers

Selected user name

Figure 7: All markers placed by a selected user in similar

entries are shown when “show markers” button is clicked.

Users can easily grasp the distribution of annotated points.

User list

Playback button

Initial scene of a selected

3D model

3D scenes of

similar entries

Marker

Figure 8: Playback of 3D Blog entries. Choose a user from

pull down menu and then click the playback button. The

system automatically shows the 3D scenes of the entries in

sequence in the 3D model viewer.

name” from the pull down menu located just below

the 3D model viewer. This menu shows the names

of those who have made entries similar to the user’s.

Second, the user chooses a function by checking ei-

ther “show markers” or “play views.”

The “show markers” function shows all the mark-

User A's Blog

User B's Blog

User C's Blog

Figure 9: Relation of similar entries. For simplicity, auto-

matic TrackBack links from user A’s entries to other users’

entries are not shown.

ers from similar entries written by the selected user, as

shown in Figure 7. Playback of the 3D Blog entries

in “show markers” mode thus helps users to visually

grasp the distribution of other user’s interests. If the

user selects the entries of another user from the pull

down menu, the user can graphically see which spots

she or he annotated.

Alternatively, the “play views” function renders

all the 3D scenes of similar entries one by one,

smoothly changing the viewpoint. This helps users

see which other users saw and annotated the 3D

model and how the viewpoint changed spatially. The

algorithm for “play views” is as follows: First the sys-

tem collects all of a particular user’s similar entries for

a particular 3D model and then categorizes them by

writer. Finally the system sorts the entries in reverse

chronological order by writer.

Figure 9 shows an example of a relation among

entries. User A wrote ﬁve entries about a particular

3D model, and three of them have automatic Track-

Back links from users B and C’s entries. Although

entry B2 has two links to user A’s blog, it is treated

as one link. In this case, user A can choose user B

or user C in addition to himself from the pull down

menu, and playback user B’s entries B1, B2, B4 or

user C’s entries C2, C3, C5 or his or her own entries

A1, A2, A3, A3, A5 in the 3D model viewer.

AUTOMATIC PLAYBACK OF 3D BLOG ENTRIES BASED ON 3D VIEWPOINT SIMILARITY

5 DISCUSSION

The most essential feature of the 3D Blog system is

that it uses 3D viewpoint information of a user. In

other words, the system places importance on from

where the user observes the 3D model as well as on

to which the user has an interest.

Considering a way of asking in our daily life, it is

quite often that we ask a question like “what is that?”

when we do not know the name of what we are look-

ing at. Those who are talking with you can under-

stand what is of interest to you since they can share

your viewpoint and follow your glance. It is, there-

fore, 3D viewpoint-based search is very natural way

to ﬁnd things of interest.

In contrast, information about what a user looking

at in the 3D Blog system is similar to a keyword in

keyword-based search. If a user uses keyword-based

search system, the user needs to know keywords or

choose keywords from a list to formulate appropri-

ate queries. This requires the user to have knowl-

edge a priori about what he/she will search while

3D viewpoint-based search does not require such as-

sumption.

The automatic playback function enables users to

take advantage of the 3D viewpoint information easily

and intuitively. As the system automatically changes

the 3D scene according to the viewpoint of entries,

users do not need to struggle with the 3D model to

recreate the 3D scene and therefore they can con-

centrate on understanding similarities among entries

written by other users.

Although the 3D Blog system has many advan-

tages, there are many topics that need further research

and development to make it fully functional.

Development of an excellent user interface for 3D

models, especially for annotating and comparing 3D

models, is crucial. It is difﬁcult, intuitively and oth-

erwise, to select an arbitrary part of a 3D model, but

this is essential when adding annotation. In addition,

easy-to-use interfaces that compare 3D models in a

single viewer should be provided since comparing ob-

jects is fundamental to research and discussion.

Functions for analysis are also important in terms

of community support tools. The “show markers”

function can provide an overview of a spatial distribu-

tion of annotated points. Users can know which areas

are of interest to community members while they can-

not know why the areas are of interest. That is func-

tions to support semantic analysis as well as physical

analysis are needed.

6 CONCLUSION

We discussed the automatic playback function for 3D

Blog entries. This function is useful especially when

viewing similar entries that the 3D Blog system au-

tomatically ﬁnds based on 3D viewpoint information

since users can easily and intuitively understand the

similarities among entries by viewing 3D scenes.

The 3D Blog system uses 3D viewpoint informa-

tion to display blog entries and to ﬁnd similar entries

and users. With this tool, people can easily discuss

a particular topic using 3D models. The automatic

playback function expands this advantage.

Although further investigation is needed on addi-

tional functions of the system to support user com-

munication and knowledge sharing via 3D blogging,

we believe 3D blogging with the automatic playback

function has great potential and can be applied to

many domains.

REFERENCES

Barbieri, T. and Paolini, P. (2000). Cooperative visits for

museum WWW sites a year later: Evaluating the ef-

fect. In Bearman, D. and Trant, J., editors, Museums

and the Web 2000, pages 173–178. Archives & Mu-

seum Informatics.

Blosxom (2006). The zen of blogging.

http://www.

blosxom.com/

. Retreived October 24, 2006.

Falk, J. H. and Dierking, L. D. (2000). Learning from Muse-

ums: Visitor Experiences and the Making of Meaning.

AltaMira Press.

Fujita, H. and Arikawa, M. (2005). A ubiquitous photo

mapping considering users’ lines of sight. In Proceed-

ings of International Workshop on Ubiquitous Data

Management (UDM2005), pages 71–77.

hatena (2007). Hatena diary keyword.

http://d.hatena.

ne.jp/keyword/

. Retreived January 9, 2007.

Kadobayashi, R. (2005). Automatic 3D blogging to support

the collaborative experience of 3D digital archives. In

Fox, E. A., Neuhold, E. J., Premsmit, P., and Wu-

wongse, V., editors, Digital Libraries: Implementing

Stragegies and Sharing Experiences, Proceedings of

the 8th International Conference on Asian Digital Li-

braries (ICADL 2005), number 3815 in LNCS, pages

109–118. Springer.

Kadobayashi, R. and Tanaka, K. (2005). 3D viewpoint-

based photo search and information browsing. In Pro-

ceedings of the 28th Annual International ACM SIGIR

Conference on Research and Development in Informa-

tion Retrieval, pages 621–622. ACM SIGIR, ACM.

lily (2005). lily.

http://lily.sourceforge.jp/

. Re-

treived October 24, 2006.

Paolini, P., Barbieri, T., Loiudice, P., Alonzo, F., Gaia, G.,

and Zanti, M. (1999). Visiting a museum together:

WEBIST 2007 - International Conference on Web Information Systems and Technologies

How to share a visit to a virtual world. In Bearman, D.

and Trant, J., editors, Museums and the Web 1999 Se-

lected papers from an international conference, pages

27–35. Archives & Museum Informatics.

ParallelGraphics (2006). Cortona VRML Client.

http://www.parallelgraphics.com/products/

cortona/

. Retreived October 24, 2006.

SixApart (2004). Trackback technical speciﬁcation.

http://www.sixapart.com/pronet/docs/

trackback_spec

. Retreived October 24, 2006.

AUTOMATIC PLAYBACK OF 3D BLOG ENTRIES BASED ON 3D VIEWPOINT SIMILARITY