3D Searching

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Global Journal of Computer Science and Technology

Volume 12 Issue 8 Version 1.0 April 2012


Type: Double Blind Peer Reviewed International Research Journal
Publisher: Global Journals Inc. (USA)
Online ISSN: 0975-4172 & Print ISSN: 0975-4350

3D Searching
By Pranav Agarwal
Pune University MAEER’s MIT COE, India
Abstract - As the number of 3D models available on the Web grows, there is an increasing need for a
search engine to help people. Unfortunately, traditional text-based search techniques are not always
effective for 3D data. The key challenges are to develop query methods simple enough for novice
users and matching algorithms robust enough to work for arbitrary polygonal models. We present a
web-based search engine system that supports queries based on 3D sketches, 2D sketches, 3D
models, and/or text keywords. We also present a web-based search engine system that supports
multimodel queries which include both text query and sketch query. This results in faster retrieval of
the result and the percentage efficiency also increases. The net result is a growing interactive index of
3D models available on the Web (i.e., a Google for 3D models).
Keywords : Search engine, sketch query, text query, multimodel query, teddy, sketch, repository.
GJCST Classification: H.3

3D Searching

Strictly as per the compliance and regulations of:

© 2012. Pranav Agarwal. This is a research/review paper, distributed under the terms of the Creative Commons Attribution-
Noncommercial 3.0 Unported License http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution,
and reproduction inany medium, provided the original work is properly cited.
3D Searching
Pranav Agarwal

Abstract - As the number of 3D models available on the Web cars, street lamps, stop signs, etc. Will he buy a 3D
grows, there is an increasing need for a search engine to help modeling tool and build them himself? Or, will he
people. Unfortunately, traditional text-based search techniques acquire them from a large repository of 3D models on
are not always effective for 3D data. The key challenges are to
the Web? We believe that research in retrieval,
develop query methods simple enough for novice users and
matching, recognition, and classification of 3D models
matching algorithms robust enough to work for arbitrary

2012
polygonal models. We present a web-based search engine will follow the same trends that can already be observed
system that supports queries based on 3D sketches, 2D for text, images, audio, and other media. An important
sketches, 3D models, and/or text keywords. We also present a question then is how people will search for 3D models.

April
web-based search engine system that supports multimodel Of course, the simplest approach is to search for
queries which include both text query and sketch query. This keywords in filenames, captions, or context. However,
results in faster retrieval of the result and the percentage this approach can fail: (1) when objects are not 9
efficiency also increases. The net result is a growing interactive annotated (e.g., “B19745.wrl”), (2) when objects are
index of 3D models available on the Web (i.e., a Google for 3D

Global Journal of Computer Science and Technology Volume XII Issue VIII Version I
annotated with inspecific or derivative keywords (e.g.,
models).
“yellow.wrl” or “sarah.wrl”), (3) when all related
Keywords : Search engine, sketch query, text query,
keywords are so common that the query result contains
multimodel query, teddy, sketch, repository.
a flood of irrelevant matches (e.g., searching for “faces”
I. Introduction – i.e., human not polygonal), (4) when relevant keywords
are unknown to the user (e.g., objects with misspelled or

O
ver the last few decades, computer science has foreign labels), or (5) when keywords of interest were not
made incredible progress in computer aided known at the time the object was annotated. In these
retrieval and analysis of multimedia data. For cases and others, a 3D search engine is needed.[1]
example, suppose you want to obtain an image of a
horse for a Power point presentation. A decade ago, you b) How to Search For 3d Models
could: 1) draw a picture, 2) go to a library and copy a We hypothesize that shape-based queries will
picture, or 3) go to a farm and photograph a horse. be helpful for finding 3D objects. For instance, shape
Today, you can simply pick a suitable image from the can combine with function to define classes of objects
millions available on the web. Although web search is (e.g., round coffee tables). Shape can also be used to
commonplace for text, images, and audio, the discriminate between similar objects (e.g., desk chairs
information revolution for 3D data is still in its infancy. versus lounge chairs). There are even instances where a
However, three recent trends are combining to class is defined entirely by its shape (e.g., things that
accelerate the proliferation of 3D models, leading to a roll). In these instances, “a picture is worth a thousand
time in the future when 3D models will be as ubiquitous words.” Our work investigates methods for automatic
as other multimedia data are today: (1) new scanners shape-based retrieval of 3D models.
and interactive tools are making construction of detailed The challenges are two-fold. First, we must
3D models practical and cost effective, (2) inexpensive develop computational representations of 3D shape
graphics hardware is becoming faster, causing an (shape descriptors) for which indices can be built and
increasing demand for 3D models from a wide range of similarity queries can be answered efficiently. In this
people, and (3) the web is facilitating distribution of 3D paper, we investigate combinations of 3D sketching, 2D
models.[1] sketching, text, and interactive refinement based on
shape similarity. We have integrated these methods into
a) Need for 3d Search Engine a search engine that provides a publicly available index
Now a days, developments are changing the of 3D models on the Web (Figure 1.1).
way we think about 3D data. For years, a primary
challenge in computer graphics has been how to
construct interesting 3D models. In the near future, the
key question will shift from “how do we construct them?”
to “how do we find them?”. For example, consider a
person who wants to build a 3D virtual world
representing a city scene. He will need 3D models of
Author : Computer Department, Pune University MAEER’s MIT COE,
India. E-mail : Pranav.pranaw@gmail.com

© 2012 Global Journals Inc. (US)


3d Searching

4) Matching: For each user query, our web server uses


its index to return the sixteen 3D models that best
match the query. Our method answers 3D shape
queries in less than a quarter of a second for our
repository; and, in practice, it scales sub-linearly
with the number of indexed models. The main
research issue at the heart of this system is how to
provide shape-based query interfaces and
matching methods that enable easy and efficient
retrieval of 3D models from a large repository. In the
following two sections, we discuss these issues in
2012

detail for different query interfaces.[1]


April

Fig. 1.1: Screenshot of our search engine


10
It allows a user to specify a query using any
Global Journal of Computer Science and Technology Volume XII Issue VIII Version I

combination of keywords and sketches (left). Then, for


each query, it returns a ranked set of thumbnail images
representing the 16 best matching 3D models (right).
The user may retrieve any of the 3D models by clicking
on its thumbnail, and/or he may refine the search by
editing the original input or by clicking on the “Find
Similar Shape” link below any thumbnail.[1]
Figure 2.1: System Organization
II. System overview for 3d models
a) Sketch Query
The organization of our system is shown in Of course, shape similarity queries are only
Figure 2.1. Execution proceeds in four steps: crawling, possible when the user already has a representative 3D
indexing, querying, and matching. The first two steps model. In some cases, he will be able to find one by
are performed off-line, while the last two are done for using a text search. However, in other cases, he will
each user query. The following text provides an overview have to create it from scratch (at least to seed the
of each step and highlights its main features: search). An interesting open question then is “What type
1) Crawling: We build a database of 3D models by of modeling tool should be used to create shapes for 3D
crawling the Web. 3D data still represents a very retrieval queries?”. This question is quite different than
small percentage of the Web, and high quality the one asked in traditional geometric modeling
models represent an equally small percentage of all research. Rather than providing a tool with which a
3D data. So, we have developed a focused crawler trained user can create models with exquisite detail
that incorporates a measure of 3D model “quality” and/or smoothness properties, our goal is to allow
into its page rank. Using this crawler, we have novice users to specify coarse 3D shapes quickly. In
downloaded 17,834 VRML models from the Web. particular, the interface should be easy to learn for first
We augment this database with 2,873 commercial time visitors to a website. Of course, this requirement
models provided by 3D vendors. rules out almost every 3D modeling tool available today
2) Indexing: We compute indices to retrieve 3D models – i.e., it would not be practical to require everybody who
efficiently based on text and shape queries. In wants to use a 3D search engine to take a three week
particular, we have developed a new 3D shape training course to learn the complicated menu structure
descriptor based on spherical harmonics that is of a commercial CAD tool. Instead, we have investigated
descriptive, concise, efficient to compute, robust to two alternatives.
model degeneracies, and invariant to rotations. The first approach is to specify shape queries
3) Querying: We allow a user to search interactively for with a simple 3D sketching tool, such as Teddy [2] or
3D models. Our system supports query methods Sketch [3]. To investigate this approach, we have
based on text keywords, 2D sketching, 3D developed a query interface in which the user creates a
sketching, model matching, and iterative refinement. simple 3D model with Teddy, and then the system
We find that methods based on both text and shape retrieves similar models (see Figure 3.1).
combine to produce better results than either one Unfortunately, our early experiences suggest
alone. that even its simple gesture interface is still too hard for

© 2012 Global Journals Inc. (US)


3d Searching

novice and casual users to learn quickly. During informal


studies, we observed that most people do not readily
understand “extrusions” and “cuts,” and they have a
difficult time getting used to rotating a 3D model to get
the proper viewpoint for modeling operations. Moreover,
only certain types of shapes can be created with Teddy.
We believe that making 3D tools even simpler would
require further constraints on the types of shapes that
could be produced. Thus, we were motivated to look for
alternate sketching paradigms.[2][3]

2012
April
11

Fig. 3.2: 2D sketch query interface

Global Journal of Computer Science and Technology Volume XII Issue VIII Version I
b) Text Query
Our system also supports searching for 3D
models by matching keywords in their textual
descriptions. To support this feature, we construct a
representative document for each 3D model. The text in
that document includes the model filename, the anchor
and nearby text parsed from its referring Web page, and
ASCII labels parsed from inside the model file. Each
document is preprocessed by removing common words
Fig 3.1: 3D sketch query interface
(stop words) that don’t carry much discriminating
Our second approach is to draw 2D shapes information, such as “and”, “or”, “my”, etc. We use the
with a pixel paint program and then have the system SMART system’s stop list of 524 common words as well
match the resulting image(s) to 2D projections of 3D as words specific to our domain (e.g. “jpg”, “www”,
objects (Figure 3.2). The main advantage of this “transform”, etc.). Next, the text is stemmed (normalized
approach is that the interface is easy to learn. All but the by removing inflectional changes) using the Porter
most novice computer users have used a 2D paint stemmer. Finally, synonyms of the filename (without the
program before, and there are no complicated viewing extension) are added using Word-Net.
or manipulation commands. Of course, the main In order to match documents to user-specified
disadvantage is that 2D images generally have less keywords or to other documents, we use the TF-
shape information than 3D models. We compensate for IDF/Rocchio method [5], a popular weighting and
this factor somewhat by allowing the user to draw classification scheme for text documents. This method
multiple 2D projections of an object in order to better assigns a similarity score based on a term’s frequency
define its shape. in the document and its inverse frequency over all
documents.

© 2012 Global Journals Inc. (US)


2012
April 3d Searching

12
(a) Text Query
Global Journal of Computer Science and Technology Volume XII Issue VIII Version I

Fig 3.3: Text interface


c) Multi-Model Query
Since text and shape queries can provide
orthogonal notions of similarity corresponding to
function and form, our search engine allows them to be
combined. We support this feature in two ways. First,
text keywords and 2D/3D sketches may be entered in a
single multimodal query. Second, text and shape
information entered in successive queries can be
combined so that a user can refine search terms
adaptively. For instance, if a user entered text keywords
in a first query, and then clicked a “Find Similar Shape”
link, the text and 3D shape would combine to form a
second query. These types of multimodal queries are
often helpful to focus a search on a specific subclass of
objects (Figure 3.4). For example, a query with only
keywords can retrieve a class of objects (e.g., tables), (b) 2D sketch query
but it is often hard to home in on a specific subclass
with text alone (e.g., round tables with a single
pedestal). Similarly, a query with only a sketch can
retrieve objects with a particular shape, but it may
include objects with different functions (e.g., both tables
and chairs). Multimodal input can combine ways of
describing objects to form more specific queries (Figure
3.4(c)).[1]

(c) Multimodal query


Fig. 3.4: Multimodal queries are often effective at
finding specific types of objects.
© 2012 Global Journals Inc. (US)
3d Searching

III. Limitation
Better 2D image matching methods: our 2D
sketching interface would be more effective with better
image matching algorithms. Sometimes users create
query sketches with interior texture and/or details (e.g.,
eyes and mouth of a human face), and our search
engine matches them with projected images containing
only boundary outlines (e.g., just the outline of the face).
For matching purposes, the interior details in sketches
are “interpreted” as boundaries of holes in projected
images, and unexpected results are sometimes returned

2012
to the user. Of course, this problem could be rectified
somewhat by providing users with instructions or
examples about how to draw their sketches.

April
New modeling tools: future 3D modeling
systems should consider integrating shape based
matching and retrieval methods into interactive 13
sketching tools. For instance, consider a 3D model

Global Journal of Computer Science and Technology Volume XII Issue VIII Version I
synthesis paradigm in which a user draws a rough
sketch of a desired 3D model and the system “fills in the
details” semi-automatically by suggesting matching
detailed parts retrieved from a large database. In such a
paradigm, the user could retain much of the creative
control over model synthesis, while the system performs
most of the tedious tasks required for providing model
detail.

IV. Conclusion
In summary, it investigates issues in building a
search engine for 3D models. The main research
contributions are: (1) New query interfaces that integrate
text, 2D sketches, 3D sketches, and 3D models. (2) We
provide a large repository of 3D models and a way to
find the interesting ones.

References références referencias


1. Thomas Funkhouser, Patrick min, Michael Kazhdan,
Joyce Chen, Alex Halderman, and David Dobkin
Princeton University, “A Search Engine for 3D
Models” ACM Transactions on Graphics, Vol. V, No.
N, 10 202002,.
2. Zeleznik, R. C., Herndon, K. P., and Hughes, J. F.
1996. Sketch, “An Interface for Sketching 3D
Scenes”, In Proceedings of SIGGRAPH 96.
Computer Graphics Proceedings, Annual
Conference Series. 163–170.
3. Igarashi, T.,Matsuoka, S., and Tanaka, H. 1999.
Teddy: “A Sketching Interface for 3D Freeform
Design”. In Proceedings of SIGGRAPH 1999.
Computer Graphics Proceedings, Annual
Conference Series., Los Angeles, CA, 409–416.

© 2012 Global Journals Inc. (US)


14
Global Journal of Computer Science and Technology Volume XII Issue VIII Version I April 2012

© 2012 Global Journals Inc. (US)


3d Searching

This page is intentionally left blank

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy