16
Need for Semantics Now models represent just the object appearance We need to represent also its Properties Roles Behaviour Services i.e. its meaning, in a human and machine understandable way The process has to be as much automatic as possible

Need for Semantics

Embed Size (px)

DESCRIPTION

Need for Semantics. Now models represent just the object appearance We need to represent also its Properties Roles Behaviour Services … i.e. its meaning , in a human and machine understandable way The process has to be as much automatic as possible. Need for Semantics. - PowerPoint PPT Presentation

Citation preview

Page 1: Need for Semantics

Need for Semantics Now models represent just the object appearance We need to represent also its

Properties Roles Behaviour Services …

i.e. its meaning, in a human and machine understandable way

The process has to be as much automatic as possible

Page 2: Need for Semantics

Need for Semantics To exploit fully the potential of of semantic

enrichment as many properties of the real world should be represented in their virtual counterpart

We need metaphors for describing objects that are similar to what humans employ

E.g. a table= a rectangular top, 4 cylindrical legs

Page 3: Need for Semantics

Shape Annotation Documenting the 3D shape with contextual

knowledge

Knowledge related to the geometry

Knowledge related to the application domain

Knowledge related to the contentstatue, base

Restoration

Fracture lines

Erosion

Page 4: Need for Semantics

Shape AnalysisCharacterization:

Evaluation of scalar functions over the surface

Structuring:

Extraction of subparts and their spatial arrangement

Segmentation:

Identification of regions having homogeneous properties (main components or features of interest)

Page 5: Need for Semantics

Ontologies for Shape :Knowledge Modelling

modelling shapes and their associated semantics using knowledge formalisation mechanisms

metadata and ontologies provide the rules for linking semantics to shape or shape parts.

Page 6: Need for Semantics

DSW by AIM@SHAPE Many ontologies, different facets Annotate whole objects

Page 7: Need for Semantics

Geometric Search Engine

Based on Geometry

and Structure

Semantic Search Engine

Based on Metadata

advanced search engines for digital shapes

Page 8: Need for Semantics

Possible scenarios for virtual worlds semantic search for reuse:

Search among models having large geometry variations (e.g. human characters and objects)

create new VHs, reuse garments, take only interesting parts

Assistive living Simulation of domestic environments with realistic

people (e.g. with disabilities), places and danger conditions

Page 9: Need for Semantics

Automatic identification of human body parts

Plumber segmentation

Tagging human body parts

Tailor characterisation

Candidate skeletal joints

Skeleton extraction

IT’S LATE!

Page 10: Need for Semantics

ShapeAnnotator: general framework

Knowledge Base

ShapeA surface mesh

DomainAn OWL ontology

CG ToolsSegmentation Plug-ins

Shape

Annotator

Expert

Abstracted ShapeA segmented mesh

Instance

Instance

Page 11: Need for Semantics

Shape annotator Framework for part-based

annotation of 3D objects with context-dependent knowledge

IT’S LATE!

Page 12: Need for Semantics

Semantic-driven segmentations

They capture semantically relevant features automatically

what is relevant in the gaming context? IMATI segmentation methods for triangle

meshes: Fitting primitives Reeb Graphs Tailor Plumber

They extract specific kind of knowledge suitable also for virtual worlds

Page 13: Need for Semantics

Fitting Primitives Hierarchical face clustering algorithm able to

recognise planes, cylinders and spheres

Page 14: Need for Semantics

Reeb Graph The topology of a shape is coded into the

Reeb graph respect to f, which captures the evolution of the level sets of f on the shape

Page 15: Need for Semantics

Skeletons and grasping regions

Plumber (and Tailor)

It defines a shape decomposition into connected components that are either tubular features or blob regions

Landmarks

Page 16: Need for Semantics

Conclusions 3D with semantics would have high impact

Easy indexing, searching and retrieval Efficient reuse and re-adaptation Realistic interactions among virtual characters and

objects Analysis, segmentation, annotation,

knowledge base Still a lot of work to do

Manual annotation No shared conceptualisation …