Download - Londra
Londra, Shape and Animation Modelling of a
3D Dog FaceAndres Adolfo Navarro Newball
Prof. Geoff Wyvill, Dr. Brendan McCane (University of Otago)Dr. Edmond Prakash (Manchester Metropolitan University)
Several human facial animation models developed in the last 30 years.
Less attention given to animal facial models.
Animal facial anatomical features are usually humanised, oversimplified, cartoonised or ignored.
Motivation
Goal
We aim to create a virtual dog head capable of displaying facial expressions.
The expressive dog Londra
Project Overview
Successfully synthesises dog facial expressions such as anger, affection, attention, fear, happiness, yawning and smelling without displaying anthropomorphic features.
A preliminary validation suggests that most expressions were recognised consistently.
Londra
A pure bottom up form of the layered approach for the bone, muscle, complementary, skin and fur layers.
Tabulated Sphere Subsets to provide a fast way to approximate collisions between objects with constrained motion.
Key contributions
Anatomical differences between humans and dogs.
Lack of anatomical and biometric information.
Scarcity of 3D digitised data from dogs.
Challenges
Charles Darwin Dog video observation Artificial pets Validation:
◦ Quantitative◦ Qualitative◦ Performance◦ Beyond
Methodology
Darwin, C. (1890). The Expression of the Emotions in Man and Animals. London:Francis Darwin ed.
Dog expressions observations
Open nostril
Close nostril
0 1,2
37
2,2
72
3,5
05
4,6
86
5,8
T
Smell
11 Dog videos analysed which complement Darwin’s descriptions
0 0,7
71
6,5
63
7,1
29
7,7
91
8,4
65
Raise ears
Twist head right
Untwist head
T
Attention
T
0 1,4
88
1,7
2,0
28
2,5
74
3,0
64
27
,04
8
Mouth
Lips
Ears
Tongue
Anger
Darwin’s anger
Anger
0.331000000000002
0.09100000000000020.174
0.2290.1730.199
0.694000000000001
0.496000000000001
0.299
0.0759999999999999
0.754000000000003
0.0760000000000005
0.885000000000001
0.7440000000000030.7570000000000040.788
3.217
1.7
0.0940000000000002
0.767000000000003
0.1900000000000020.221
0.453
0.846000000000001
0.283
2.277
Background
Background
Background
Description of expressions
AU Name AU Name
123456789101112
Upper lip raiserLower lip depressorNostril dilatorMouth corner moverLower eyelid depressorUpper eyelid depressorEar advancerEar lowererJaw raiserTongue retractor, drawerTongue depressorEyeball mover - rotator
131415161718192021222324
Eyeball retractorEyeball vertical mover – rotatorHead raiserHead lateral moverHead rotatorTail raiserTail extenderTail lateral moverBody raiserLeft paw raiserRight paw raiserHair raiser
Anger:Tail is erect and rigid 18 (1) +
Ears are directed forward 7 (1) +
Upper lip is raised 1 (1) + 4 (0.5)
Architecture
Setting up
DFACSAU Name AU Name
123456789101112
Upper lip raiserLower lip depressorNostril dilatorMouth corner moverLower eyelid depressorUpper eyelid depressorEar advancerEar lowererJaw raiserTongue retractor, drawerTongue depressorEyeball mover - rotator
131415161718192021222324
Eyeball retractorEyeball vertical mover – rotatorHead raiserHead lateral moverHead rotatorTail raiserTail extenderTail lateral moverBody raiserLeft paw raiserRight paw raiserHair raiser
Processing
DFACSAU Name AU Name
123456789101112
Upper lip raiserLower lip depressorNostril dilatorMouth corner moverLower eyelid depressorUpper eyelid depressorEar advancerEar lowererJaw raiserTongue retractor, drawerTongue depressorEyeball mover - rotator
131415161718192021222324
Eyeball retractorEyeball vertical mover – rotatorHead raiserHead lateral moverHead rotatorTail raiserTail extenderTail lateral moverBody raiserLeft paw raiserRight paw raiserHair raiser
F
Output
Import skull
Import nose Place deformersCreate tongue
Place eyes
Place muscles
Smoothed skinwith lips
TSSs for:-Jaw motion-Muscle and skin interaction-Tongue motion
Euler integration
Deformed skinDeformed tongueDeformed noseTransformed eyes
Render animated expression
Render with OpengGL
Render with Blender
Bottom up approach
Bone layer Muscle layer
Temporalis
Zygomaticus
FrontalisLevator Oculis
Masseter
Digastricus
Orbicularis OrisMentalis
LevatorAuricularisrostrales
Orbicularis OculisAuricularisdorsales
Barr, A. H. (1981). Superquadrics and Angle-Preserving Transformations. IEEE Comput.Graph. Appl., 1 (1), 11–23.
α DistEyes
P
D
B)A)
X ≤ X0 X ≥ X0
Sn (X0, Y0, Z0)Z
Cornea
Sclera
Iris
Pupil
Back Middle Tip
C)
D) E)
GrooveMT
MTM
D
DistEyes
2sin 1
Complementary layer
A) B) C)
R
R +d
αAxis
D0
D0
D1
D2D) E)
Skin Layer
Skin layer -Deformation
P
P’
R
F
P
P’
R
F Nm
Ns
D
R F
Muller, M., Heidelberger, B., Teschner, M., and Gross, M. (2005). Meshless deformationsbased on shape matching. In SIGGRAPH ’05: ACM SIGGRAPH 2005 Papers,New York, NY, USA, 471–478. ACM.
King, S. A. (2001). A facial model and animation techniques for animated speech. Ph.D. thesis, The Ohio State University.
Kobbelt, L. (2000). √3-subdivision. In SIGGRAPH ’00: Proceedings of the 27th annualconference on Computer graphics and interactive techniques, New York, NY, USA,103–112. ACM Press/Addison-Wesley Publishing Co.
Tabulated Spheres Subsets
Definition
Msa Ms
b
a b
Line of motion, 1 DOF Ms= Approximation with spheres.n = number of spheres in a = 9m = number of spheres in b = 18 n X m = 9 X 18 = 162 tests
Msc= Colliding spheres subset.n = 3m = 7n X m = 21 tests
Msca
a b
Mscb
Mssa
a b
Mss= Minimal subset.n = 3m = 3n X m = 9 testsTSSAB = Msa X Msb = compares non redundant spheres of the same colour against spheres of the same colour only.TTSS = 3 tests
Mssb
22 tests time step
Jaw motion
Muscle skin interaction
WANG W., WANG J., KIM M.-S.: An algebraic conditionfor the separation of two ellipsoids. Comput. Aided Geom.Des. 18, 6 (2001), 531–539.
The use of a more freely moving object. A flexible object interacting with more than
one object The use of an object which has been divided
in several sections which need to interact with each other.
Extraordinary cases where some collisions need to be ignored.
Tongue motion
A)
α
β
4α
4α
2α
4β
4β 2
β
B) C)
2 – 5 tests
Expression synthesis
DFACS
AU13: Raise eyesAU12: Left eyes AU6: Raise ears
AU11: Twist head AU9: Raise head
AU -14: Relax eyesAU -9: Lower head AU9: Raise headAU7: Open mandible AU16: Lick
A)
B)
C)
D)
Original
Anger Attention Affection II Yawn
Smell
Happiness Fear Affection I
Londra’s videos can be downloaded at: http://cic.javerianacali.edu.co/~anavarro/Londra/
Validation
U flow V flow Mixed flowTwo framesA)
A)
B)
C)
Survey
Performance
60 145 245 370 10510
5
10
15
20
25
18.517.2
14.7 13.2 13.7
21.3 20.819.2
17.9 18.2
FPS, SL2
FPS, SL1
Collision tests
FPS
Beyond
We followed a new pure bottom up approach which starts with a skull and does not require a pre-existing facial mesh.
We introduced TSSs, a fast and appropriate method for constrained object interaction.
We validated eight of Londra’s expressions successfully.
Conclusions
Enhancing our bottom up approach by creating an anatomically accurate skull reshaping method in conjunction with zoology. Then, automating muscle placement.
Our TSSs open a full field of research for constrained object interaction. For example, mirrored TSSs.
The Londra model could be expanded to other real or non real non human creatures. And some of the techniques could be used in human systems.
Further work
Some videos
• Blender 1• Blender 2• Blender 3