Creating life in Ancient Sites

Transcription

1 Creating life in Ancient Sites Presentation to Dongguk University Nadia MAGNENAT-THALMANN Switzerland Background Human body has been a long time inspiration topics 1

2 Background Generating human models I How to: build geometry to represent human bodies? Methods: by interactive design Surface model, multi-layered by anatomical models physics-based technology Accurate sumulation of muscles, bones and fat tissues reconstruction techniques Capture shapes in real world From 2D or 3D data [Wilhelms97] [Shen [Schroeder92] 95] [Hoppe94] [Heckbert94] [Carr01] [Nedel Commercial and Thalmann98] softwares scanners [Hilton99][Lee00][Sand03] [Lee et al 00] [Aubel et al 02] [Shen95] [Sand et al 03] Background 3D image capture technology - Why? Systematic observation of human bodies. Example: CAESAR (Civilian American and European Surface Anthropometry Resource) project 3D whole body scanners (c.f. tape measure) extract body measurements 3D Anthropometric database Advantages: Accuracy Speed Comfort 2

3 Background 3D image capture technology - Limitations Classical problems Complexity, noise removal, hole-filling Courtesy by Clive Arrowsmith Other requirements body model of desired sizes standards, individuals easily and almost instantly Not quite possible solely by using the capture technology!! Jung-Ang daily manufacture s customers expectation: : Show me a my 3d mannequin 3d model without such and such bothering size me too much!! Background Generating human models II How to: systematically obtain a VARIETY of human shapes? (cf. one model at a time) Methods: by segment scale by variational modeling by using examples Trade-offs between controllability quality of shape time costs [Azuola et al 94] [DeCarlo et al 98] [Blanz and Vetter99] 3

4 Face Cloning Face Cloning: Input: 2 Photographs. Generic head & animation. Method: Feature based. Output: Animatable virtual human. Real photos of Guido Virtual Guido The texture mapping Creating textured cloth body Body Cloning & textured cloth: Input: Three photographs. H-Anim 1.1 generic body. Feature: Edge based. Output: Animatable virtual human. 4

5 Textured cloth body animation Motion capture: Optical motion capturing system (Vicon) Anatomical converter > sensor information in model animation parameters. Motion Capture Textured cloth body animation Some examples 5

6 Research in Mechanical Models for Cloth Current models focus either on realism, Finite elements Continuum mechanics Or speed. Particle systems Implicit integration 6

7 Early Cloth Simulation Models B. Lafleur, N. Magnenat-Thalmann, D. Thalmann, Cloth Animation with self-collision detection, Proc. IFIP Conf. On graphics modelling, Tokyo, 1991 Viscoelastic surfaces using Lagrange equations Carignan, Yang, Magnenat-Thalmann, dressing animated synthetic actors with complex clothes, Proc. SIGGRAPH 92 Modified Terzopoulos model with octree collision detection and advanced pattern-seaming garment design General Mechanical Parameters Internal Forces (From surface deformations) Elasticity (metric, curvature) Viscosity Plasticity External Forces (From environment interactions) Gravity, Air Viscosity Contact reaction, Friction Miscellaneous Interactions 7

8 Mechanical Simulation Systems Particle Systems Discrete representation of the system using a set of masses interacting with forces modeling macroscopic mechanical behavior. Spring-Mass Systems Relevance of mechanical parameters Identify the relevance of mechanical parameters in the motion of fabric: Garments in motion: Dissipative parameters have importance (viscosity, plasticity, aerodynamic interactions). Dissipative parameters are not measured by standard experiments (FAST, KES, ). 8

9 Simulated Mechanical Parameters Elasticity: stress/strain curve. Viscosity: stress/(δstrain/δt) curve. Poisson coefficient Aerodynamic effects: force dependent on the relative speed of the air. Plasticity: hysteresis in stress/strain curve. Gravity Material (internal) parameters Metric elasticity: measurement of the fabric elongation elasticity (N.m-1) Weft and Warp elasticity: elasticity along the Weft and Warp directions Shear elasticity: elasticity for a shearing deformation between weft and warp directions Bending elasticity: measurement of the fabric bending elasticity (N.m) Weft and Warp bending: bending along the Weft and Warp directions Viscosity parameters: defined for each elastic parameter Density: mass per surface unit of the fabric (Kg.m-2) 9

10 Contact & Environment parameters Thickness of the fabric (m) Coulombian friction: ratio between the maximum tangential contact force and the normal pressure force between two surfaces in contact Gravity: nominal acceleration of objects left at rest (9.81 m.s-2) Aerodynamic viscosity: force exerted on a fabric per surface unit and per velocity unit between the fabric speed and the air speed: wind Normal (Flowing: N.m-3.s) and tangential (Damping: N.m- 2.(m.s-1)-1) components relative to the orientation of the fabric surface Example of material parameters Weft Elasticity N.m -1 Weft Elasticity 25 N.m -1 Weft Elasticity 50 N.m -1 Warp Elasticity N.m -1 Warp Elasticity 25 N.m -1 Warp Elasticity 50 N.m -1 Shear G 60 N.m -1 Shear G 86 N.m -1 Shear G 55 N.m -1 Weft Bending N.m Weft Bending N.m Weft Bending N.m Warp Bending N.m Warp Bending N.m Warp Bending N.m Density Kg.m -2 Density Kg.m -2 Density Kg.m -2 Cotton Tencel Linen 10

11 The making of clothes with our Fashionizer platform DEMO 2D patterns 3D garments from 2D patterns 11

12 «Prêt-à-porter» Collection Garment Creation Tools Interactive tools for creating, draping, animating and prototyping garments on animated virtual characters [1] [1] P. Volino, N. Magnenat-Thalmann, Accurate Garment Prototyping and Simulation, Computer-Aided Design and Applications journal, CAD Solutions, 2(5), pp ,

13 Research collaboration with Adidas Garment Prototyping Accurate Simulation of Parameters Evaluation of strain, stress and contact pressure all over all over the cloth surface. 13

14 Color coded visualizations of deformation parameters Deformation in Warp direction Deformation in Weft direction Deformation in Shear direction Color coded visualizations of deformation parameters Pressure 14

15 Haute Couture Garments Mode, Passion et Collection - Le regard d une femme Musée d Art et d histoire, Genève Haute Couture Garments Lacroix Lanvin Nina Ricci Paco Rabanne 15

16 Haute Couture Garments Exhibition Robert Piguet/Yverdon Modèle designed by Bohan 16

17 Exhibition Robert Piguet/Yverdon Modèle designed by Givenchy Exhibition Robert Piguet/Yverdon Modèle designed by Givenchy 17

18 Open Problems in Hair Simulation Hair Shape Modeling Many recent attempts, however not yet matured Hair Dynamics Stiffness dynamics of individual hair was grossly approximated considering current computing power (1GHz, 1GB RAM) Hair-hair interaction not attempted until recently Open Problems in Hair Simulation Hair Rendering Fairly matured and is available through commercial systems. However, global-illumination for hair with detailed shading models is not yet attempted. Real-time hair animation and rendering With possible compromise on the realism Hair interaction with a user (hairdresser for example) 18

19 Our Non Real-Time Approach Hair Shape Modeling* Hair shape as streamlines of fluid flow Hair Rendering Single iteration hair rendering including volumetric shadows using graphics hardware *Sunil Hadap and. "Interactive Hair Styler based on Fluid Flow", Eurographics Workshop on Computer Animation and Simulation'2000, Interlaken 2000 Making of Placement of smoothed particles to realize hair-fluid 19

20 Our Non Real-Time Approach Hair Dynamics** Stiffness dynamics of non-straight hair strand Elaborate inertial dynamics using reduced coordinate formulation Hair-hair, hair-body and hair-air interaction dynamics **Sunil Hadap,, "Modeling Dynamic Hair as a Continuum", Computer Graphics Forum, Volume 20, Issue 3, Eurographics 2001 Proceedings, Manchester, United Kingdom, September 2001 Individual Hair Dynamics Body forces Without stiffness Collision as fluid boundary With stiffness 20

21 Results Computer Animations Bunch of hair Virtual Experimental Setup (collaboration with Unilever) Switch length 24cm Number of hair strands 1000 Hair segment length 0.9cm Amplitude 4cm Frequency 1Hz Number of cycles - 2, 21

22 Variations in Strand Stiffness Increase in stiffness gives considerable phase difference. In reality, it is the change in hair thickness which varies the bending stiffness. However, it is hard to model various hair thickness We may simulate the process by changing the stiffness coefficient E Observe the tip movement with respect to the movement of the base E E=250 E=500 E=750 Simulation - Adjusting the air drag Normal hair makes 5-6 oscillations before coming to rest. When hair is treated with styling products, air drag increases to reduce number of oscillations to merely

23 Variations in Internal Damping The internal damping is associated to energy loss as elastic material undergoes deformation Though the large change in internal damping has minute effect on the hair movement, it is important to increase stability of the simulation ID ID=0.1 ID=0.2 ID=0.3 Variations in Hair-hair Interaction The spread in the bunch of hair is a result of hair-hair interaction One can vary the constant Kc to change this interaction An immediate effect is on the spread of the tip One can also observe how the tip makes a bouncy movement as Kc increases. K c Equation of state K=200 K=1000 K=5000 K c ρ0 ρc 23

24 Results Computer Animations Hair fall Results Computer Animations Blown by wind 24

25 Result Real-Time Animation of Hairstyles Managing Complexity of Hair Simulation More than hair strands Mechanical behavior of individual strands Collisions and friction on the skull Collisions and friction between strands Physical state (wetness, styling products) Aerodynamics No chance to simulate an explicit mechanical model in real-time. 25

26 Real-Time Animation of Hairstyles Ideas for Speeding Up Hair Simulation Simplification of the mechanical model Reduction of the degrees of freedom Macroscopic approximations Interpolation Techniques Wisp or Cluster models Particle systems of variable topology Volume hair models Free-Form Deformations 1. Real-Time FFD Hair Animation(Research with Kao Corporation, Japan) The main idea: Deforming the complete hairstyle using a mechanically-animated Free-Form Deformation lattice. 26

27 1. Real-Time FFD Hair Animation Main advantages: Drastic and controllable reduction of the number of degrees of freedom of the mechanical model => Real-time mechanical simulation Simple and fast interpolation scheme for computing the deformed hairstyle => Real-time motion of any feature of the hairstyle Any hairstyle can be animated => Versatility and design simplicity 2. FFD Hairstyle Deformation Building the lattice around the hairstyle The lattice is attached to the rigid-body motion R of the skull The lattice is then deformed by mechanical simulation 27

28 3. Animating the FFD Lattice Mechanical Properties to be Simulated Mechanical behavior of hair: Density, Elasticity Interactions between strands External forces exerted on hair: Gravity Aerodynamic effects Collisions between hairs and other objects: Body parts: Skull, shoulders Other objects 3. Animating the FFD Lattice Collision Effects Between the hair and the body Head Shoulders Metaball-based model of body parts Approximate modeling with a low number of primitives Easy animation of body deformations Attached to the rigid head motion (skull) or the body skeleton (shoulders) 28

29 3. Animating the FFD Lattice Numerical Integration Particle system integrated by using the implicit Inverse Euler method Using the Conjugate Gradient as linear system solver On-the-fly matrix assembly for capturing all the nonlinearities of the mechanical system Good robustness with any mechanical system when using constant time steps 4. Scalability and Level-of-Detail Options for Adjusting the Accuracy Tradeoff Number of lattice attachments The more attachments, the more accurate the mechanical behavior of the hairs Size of the free-form deformation grid The finer the grid, the richer and diverse the deformation patterns of the hair Number of metaballs The more metaballs, the more accurate the collisions between the hair and the body 29

30 4. Scalability and Level-of-Detail Performance and Level-of-Detail Level-of-Detail schemes have to be combined with adequate multiresolution rendering techniques 4. Scalability and Level-of-Detail Benefits of the FFD Model Scalability, for efficient LOD implementations Compatible with any hair rendering method Weak reliance on the strand nature of hairs Total freedom of hairstyle design using any standard design tool 30

31 CULTURAL HERITAGE: European Research Project Lifeplus: augmented life in Pompei LIFEPLUS proposes new Augmented Reality narrative spaces for the innovative revival of life in ancient frescospaintings in Pompeii 31

32 Scenario 3, e.g. La Villa dei Misteri Scenario 1, e.g. Thermopolio di Vetuzio Placido Scenario 2, e.g. La Casa dei Vettii Research in CLOTHES AND FACIAL ANIMATION Real-time Cloth and hair simulation Hybrid deformation Real-time Facial emotion expression and Speech animation 32

33 33

34 movie 34

35 Research in HUMANS Simulation Real-time realistic skin rendering and interactive programmable shading module movie Artificial life methods for behavioural animation of virtual characters 35

36 Early Mixed Realities simulations in ancient Pompeii (offline AR) movie 36

37 AR Life: User Experience 73 User arrives at the AR Life hot spot inform about AR Life Simulation option start theater-like immersive AR simulation User participation and interaction using: HMD + head mounted camera watching AR revival of the ancient life dramatic scenario featuring high quality Virtual Humans blended into the real scenery moving around the scene, changing point of view AR Life: Functional Elements 74 main functional elements real-time camera tracking: marker-less real-time VR simulation: advanced Virtual Humans blending track camera blend Camera and VR result generate VR 37

38 38

39 Making of 39