canim-1 - The University of Texas at Dallas

meanchildlikeMechanics

Oct 31, 2013 (3 years and 11 months ago)

77 views

Multimedia System and Networking Lab @ UTD Slide
-

1



University of Texas at Dallas B. Prabhakaran

Computer Animation


Main Text:


"Computer Animation: Algorithms & Techniques“,
Rick Parent, Morgan Kaufman publishers.


"3D Computer Graphics: A Mathematical
Introduction with OpenGL", Samuel R. Buss,
Cambridge University Press

Multimedia System and Networking Lab @ UTD Slide
-

2



University of Texas at Dallas B. Prabhakaran

Course Outline

1.
Skeletons

2.
Quaternions

3.
Skinning

4.
Facial Animation

5.
Advanced Skinning

6.
Channels & Keyframes


7.
Animation Blending

8.
Inverse Kinematics

9.
Locomotion

10.
Particle Systems

11.
Cloth Simulation

12.
Collision Detection

13.
Rigid Body Physics


Multimedia System and Networking Lab @ UTD Slide
-

3



University of Texas at Dallas B. Prabhakaran

Contact Information

B. Prabhakaran

Department of Computer Science

University of Texas at Dallas

Mail Station EC 31, PO Box 830688

Richardson, TX 75083

Email:
bprabhakaran@utdallas.edu

Fax: 972 883 2349

URL: http://www.utdallas.edu/~praba

Phone: 972 883
4680

Office:
ECSS 3.706

Office Hours:
Tuesdays/Thursdays 10.30
-
11.15am



Other times by appointments through email

Announcements: Made in class and on course web page.

TA: TBA.

Multimedia System and Networking Lab @ UTD Slide
-

4



University of Texas at Dallas B. Prabhakaran

Prerequisites


CS 2236 (CS2), CS/SE 3345 (Data structs &
Alg. analysis), Math 2418 (Linear Algebra).


Familiarity with


Vectors (dot products, cross products…)


Matrices (4x4 homogeneous transformations)


C++ or Java


Object oriented programming


Basic physics

Multimedia System and Networking Lab @ UTD Slide
-

5



University of Texas at Dallas B. Prabhakaran

Evaluation


1 or 2 Homeworks


1 Final Exam: 75 minutes or 2 hours (depending on class room
availability). Mix of MCQs and Short Questions.


Programming Projects

Multimedia System and Networking Lab @ UTD Slide
-

6



University of Texas at Dallas B. Prabhakaran

Grading


70% Projects


5% Homeworks


25% Final

Multimedia System and Networking Lab @ UTD Slide
-

7



University of Texas at Dallas B. Prabhakaran

Schedule


Final Exam: Last week of class OR As per UTD schedule


Projects and homework(s) schedules will be announced in class and
course web page, giving sufficient time for submission.


Multimedia System and Networking Lab @ UTD Slide
-

8



University of Texas at Dallas B. Prabhakaran

Programming Projects


No copying/sharing of code/results will be tolerated. Any instance of
cheating in projects/homeworks/exams will be reported to the University.


No copying code from the Internet
.


2 individual students copying code from Internet independently: still
considered copying in the project !!


Individual projects.


Deadlines will be strictly followed for projects and homeworks
submissions.


Projects submissions through eLearning.


Demo may be needed


Multimedia System and Networking Lab @ UTD Slide
-

9



University of Texas at Dallas B. Prabhakaran

Cheating


Academic dishonesty will be taken seriously.


Cheating students will be handed over to Head/Dean
for further action.


Remember: home works/projects (exams too !) are to
be done individually.


Any kind of cheating in home works/ projects/ exams
will be dealt with as per UTD guidelines.


Cheating in any stage of projects will result in 0 for the
entire set of projects.


Multimedia System and Networking Lab @ UTD Slide
-

10



University of Texas at Dallas B. Prabhakaran

Proposed Projects


3 projects


Encourage you to come up with your own project proposal too


Announcements will be made soon


Use OpenGL


Possible use of Autodesk 3D Max


Or other public domain software


C/C++ mostly


Multimedia System and Networking Lab @ UTD Slide
-

11



University of Texas at Dallas B. Prabhakaran

Applications


Special Effects (Movies, TV)


Video Games


Virtual Reality


Simulation, Training, Military


Medical


Robotics, Animatronics


Visualization


Communication

Multimedia System and Networking Lab @ UTD Slide
-

12



University of Texas at Dallas B. Prabhakaran

Computer Animation


Kinematics


Physics (a.k.a. dynamics, simulation,
mechanics)


Character animation


Artificial intelligence


Motion capture / data driven animation

Multimedia System and Networking Lab @ UTD Slide
-

13



University of Texas at Dallas B. Prabhakaran

Animation Process

while (not finished) {

DrawEverything();

MoveEverything();

}




Interactive vs. Non
-
Interactive


Real Time vs. Non
-
Real Time

Multimedia System and Networking Lab @ UTD Slide
-

14



University of Texas at Dallas B. Prabhakaran

An Example

Multimedia System and Networking Lab @ UTD Slide
-

15



University of Texas at Dallas B. Prabhakaran

The process involves…


Motion Capture (Data Acquisition)

Multimedia System and Networking Lab @ UTD Slide
-

16



University of Texas at Dallas B. Prabhakaran

Human Motion Capture

Multimedia System and Networking Lab @ UTD Slide
-

17



University of Texas at Dallas B. Prabhakaran

UTD’s Motion Capture Facility…

Multimedia System and Networking Lab @ UTD Slide
-

18



University of Texas at Dallas B. Prabhakaran

Captured 3D Motion: E.g., 1

Multimedia System and Networking Lab @ UTD Slide
-

19



University of Texas at Dallas B. Prabhakaran

Captured 3D Motion: E.g., 2

Multimedia System and Networking Lab @ UTD Slide
-

20



University of Texas at Dallas B. Prabhakaran

Motion Capture Matrix

Frame

pelvis<A
-
X>

pelvis<A
-
Y>

pelvis<A
-
Z>

pelvis<T
-
X>

pelvis<T
-
Y>

pelvis<T
-
Z>

1

-
4.62953

-
36.2313

176.458

590.269

166.422

797.569

2

-
4.65407

-
36.2417

176.453

590.039

166.612

797.706

3















Pelvis Joint Data:

Multimedia System and Networking Lab @ UTD Slide
-

21



University of Texas at Dallas B. Prabhakaran

Applying Motion Data to 3d Model

Multimedia System and Networking Lab @ UTD Slide
-

22



University of Texas at Dallas B. Prabhakaran

Animated 3D Model

Multimedia System and Networking Lab @ UTD Slide
-

23



University of Texas at Dallas B. Prabhakaran

Animation: Applications &
Possibilities


Using an Expert to Train


Animation Toolkit


Content Based Retrieval of 3D Models & Motions


Networked 3D Games


Streaming 3D Models and Motions


Copyright / Content Protection


Collision Detection

Multimedia System and Networking Lab @ UTD Slide
-

24



University of Texas at Dallas B. Prabhakaran

Application: Improve Your Game !

Multimedia System and Networking Lab @ UTD Slide
-

25



University of Texas at Dallas B. Prabhakaran

3D
novice

motion & 2D
expert

motion



We can get novice pitching data using motion capture system


There are bunch of videos include expert’s pitching motion


2D video data has expert’s stylistic actions


3D motion data compensate for the incompleteness of 2D data

Multimedia System and Networking Lab @ UTD Slide
-

26



University of Texas at Dallas B. Prabhakaran

2D Motion Analysis

position

time

Motion analysis by tracking the object (e.g. right hands)

Multimedia System and Networking Lab @ UTD Slide
-

27



University of Texas at Dallas B. Prabhakaran

Constraint of 2D motion Analysis

3D motion capture data

2D video motion analysis data

• We need to compare the dissimilarity between 2D & 3D data

• However, 2D data from a single camera doesn’t have enough information


for comparing with 3D motion capture data


Multimedia System and Networking Lab @ UTD Slide
-

28



University of Texas at Dallas B. Prabhakaran

Reconstruction 3D from 2D

using HMM


Calculate most probable style
-
path given 2D
observations



Argmax P(Q|O1O2…OT)




Red
: 3D novice motion data

Blue
: reconstructed 3D motion data

Multimedia System and Networking Lab @ UTD Slide
-

29



University of Texas at Dallas B. Prabhakaran

Resynthesis


Following to the
reconstructed 3D expert
style.


Red
: 3D novice motion data

Blue
: reconstructed 3D motion data

Multimedia System and Networking Lab @ UTD Slide
-

30



University of Texas at Dallas B. Prabhakaran

Another Fun
: 3D Tennis Game

Realistic Tennis game


Topspin (EA Sports)


Multimedia System and Networking Lab @ UTD Slide
-

31



University of Texas at Dallas B. Prabhakaran

Application of learning expert style
motion to 3D Tennis Games



Tennis novice can learn by comparing style
with realistic professional player



Motion capture system can capture novice’s
naïve actions (serve, swing, volley ..)



We can build realistic professional expert’s
actions by motion resynthesis (3D motion
reconstruction from 2D video data)



Multimedia System and Networking Lab @ UTD Slide
-

32



University of Texas at Dallas B. Prabhakaran

Parallel FSM (Finite State Machine)


Motion capture data is not high
-
level semantic
data (sequences, not segmented data)


To identify “high
-
level action”, we prepare
neural network and Parallel FSM


Parallel FSM is needed since human actions
happened not in a separate way



Multimedia System and Networking Lab @ UTD Slide
-

33



University of Texas at Dallas B. Prabhakaran

Behavior Modeling: Neural Network
& Parallel FSM


Sensor layer : two input nodes which notice the object’s movement


& boolean value of range respectively.

• Control layer : works as a hidden layer

• Stand, Straight and Grab nodes (output nodes) also initial states of each


FSMs.

Multimedia System and Networking Lab @ UTD Slide
-

34



University of Texas at Dallas B. Prabhakaran

High
-
level behavior recognition using
Motor
-
graph


To interpret low
-
level actions to high
-
level behaviors

• Motor
-
graph is designed with states of FSMs

• Nodes : each state , edges: state transitions

• (c) subsumed by (b) sub
-
graph

Multimedia System and Networking Lab @ UTD Slide
-

35



University of Texas at Dallas B. Prabhakaran

Translate into “serve” action by
Parallel FSM & Motor Graph

Locomotion FSM

Head FSM

Arm Hands FSM

A0

H0

L0

A1

A2

L1

L2

A4

A5

H1

H2

H3


Neural Network sensors the participant’s action and hand it to FSMs

• Each FSM recognize the state
-
transition and draw it to motor
-
graph

• This action motor
-
graph is subsumed by “
serve

minimum motor graph



translate this action as “
serve
” !!


Multimedia System and Networking Lab @ UTD Slide
-

36



University of Texas at Dallas B. Prabhakaran

System Architecture:

Analysis novice’s style & feedback expert
-
like
action


Behavior Translation

Neural Network & FSM

Motor Graph

Hands style

Head style

Locomotion style

Style Analysis

Matched expert’s serve

Showing a developed

serve with user’s style

Novice

s behavior

Multimedia System and Networking Lab @ UTD Slide
-

37



University of Texas at Dallas B. Prabhakaran

Animation Toolkit


Animation Authoring
Through Reuse:


Motion mapping


Inverse kinematics


Example:


GET walking FROM
Andy


GET waving FROM
Andy


JOIN Andy.walking
WITH Andy.waving

Multimedia System and Networking Lab @ UTD Slide
-

38



University of Texas at Dallas B. Prabhakaran

Animation Authoring Toolkit

Multimedia System and Networking Lab @ UTD Slide
-

39



University of Texas at Dallas B. Prabhakaran

Animation Query Handling


Partial Fuzzy Query Resolution:


Primary attribute

centric query resolution


insert an animation sequence where
Mickey Mouse

is walking slowly
in a park with a fountain or a statue in the background


Heuristics for retrieving top k objects


Maximal grade based approach


Maximal attributes based approach


Threshold algorithm


Decent precision and recall shown.


“Partial Fuzzy Query Resolution for Animation
Authoring” (Phanivas Kotharu, MS Thesis, UTD).

Multimedia System and Networking Lab @ UTD Slide
-

40



University of Texas at Dallas B. Prabhakaran

Network

Capture

Indexing

………..

………..

………..

Index Tree

Metadata based Query

Query by Example

Compression

Query / Data
Processor

Deliverable Data

Animation Toolkit

Multimedia System and Networking Lab @ UTD Slide
-

41



University of Texas at Dallas B. Prabhakaran

Shape Analysis of 3d models


Applications


-
Categorization

of shapes

-
Indexing

techniques of 3d models

-
Querying
techniques for 3d model database


Ultimate goal:

Content based 3d model search

Multimedia System and Networking Lab @ UTD Slide
-

42



University of Texas at Dallas B. Prabhakaran

Streaming 3D Games Over the Internet

3D Streaming

Server

Rendering Client

Network

Multimedia System and Networking Lab @ UTD Slide
-

43



University of Texas at Dallas B. Prabhakaran

3D Model Streaming


Advantages:


1 Base Mesh + M
Refinements

=
Original Mesh


Bandwidth Friendly


Drawbacks:


Intolerant to
Transmission errors


Not friendly for
Real Time 3D
Streaming


Base Mesh

Faces: 4281

Vertices: 2249

Size:131KB

Mesh after 5 Batches


Faces: 23675

Vertices: 11946

Size: 748KB

Original Mesh

Faces:
69451

Vertices: 35947

Size: 3MB

1.

2.

3.

Multimedia System and Networking Lab @ UTD Slide
-

44



University of Texas at Dallas B. Prabhakaran

Content Protection of 3D models and
MoCap Data


3D models and MoCap Data


-

Commercial value (“money”)


-

Requires lot of human effort


Tampering and piracy of data:


loss of information, with ultimate

loss of time
and money
.


Faulty training & education


How do we do content protection to avoid
piracy and tampering ?

Multimedia System and Networking Lab @ UTD Slide
-

45



University of Texas at Dallas B. Prabhakaran

Tamper Proofing Game Data

Secure data used for driving the game (different from
outcome data)


Tamper proofing


Detect (and possibly correct) attacks on data


Water marking (more to do with copyrighting)


Focus both on 3D models, motion, apart from other
data



Multimedia System and Networking Lab @ UTD Slide
-

46



University of Texas at Dallas B. Prabhakaran

Collision Detection



Authoring operations may lead to unintentional collisions.




Collision detection: alert authors on possible collision
detection and suggest possibilities for avoiding them.




Previously used approaches for Collision Detection can be
classified into 3 categories.



Cell Based



Tree Based



Bounding Object Based






Multimedia System and Networking Lab @ UTD Slide
-

47



University of Texas at Dallas B. Prabhakaran

Cell
-
based Approach



D
ivide the entire search space into a number of cells and a


collision possibility is triggered if two objects come in one


cell.




Disadvantages:




High memory usage.




Tough to correctly determine the size of each cell.




Too small a cell: objects occupying many cells and hence more
collision tests.




Too big a cell: unnecessary tests being carried out.





Multimedia System and Networking Lab @ UTD Slide
-

48



University of Texas at Dallas B. Prabhakaran

The motion of Object A causes a rippling effect on
Object C after colliding with Object B. The possibility of
Collision of Object C can be detected when the bounding
sphere of Object A encompasses C. This helps in Early
detection of the Collision.

Multimedia System and Networking Lab @ UTD Slide
-

49



University of Texas at Dallas B. Prabhakaran

Course Outline

1.
Skeletons

2.
Quaternions

3.
Skinning

4.
Facial Animation

5.
Advanced Skinning

6.
Channels & Keyframes


7.
Animation Blending

8.
Inverse Kinematics

9.
Locomotion

10.
Particle Systems

11.
Cloth Simulation

12.
Collision Detection

13.
Rigid Body Physics