[x3d-public] [h-anim] Basic Animataion data

[Joe D Williams]

----- Original Message ----- 
From: "doug sanden" <highaspirations at hotmail.com>
To: <x3d-public at web3d.org>
Sent: Friday, November 11, 2016 7:40 AM
Subject: Re: [x3d-public] [h-anim] Basic Animataion data


> Looks good - quats.
> -Doug
> ..
> And I think if dropping general transform -with scales- for joints, 
> that means if you have skinNormals, you do the same transform on 
> normals (as coords) instead of inverse-transpose, which IIRC 
> corrects for anisotropic-scale-induced shear.

That sounds big. I think you would need to write a fairly complete 
outline on what to change.
Put this in as a spec comment. I have one in there that asks for 
inclusion of either axis-angle of unit quaterions.
Of course the main issue, is that as far as I know, collada and glTF 
will only use unit quats for animation sequence transposrt.



> ...
> freewrl converts axis angles to quats internally for some things 
> already - to avoid the pole singularity with axis angles.

Great, now will you implement the HAnim stuffs.


> ...
> Would like to see more goodies in HAnim / Anim.
> When implementing hanim in freewrl I was shocked by how little work 
> the web3d browser does, and how much work is left for the content 
> developer to do.

Notice that the Joint node does have fields for data to do IK, so that 
a joint could be moved by other than hierarcal joint rotations. Just 
that as far as I know, those fields never got implemented but I guess 
with hope, the fields were left in the X3D spec. .


> For example IK. Q. Should we have a general IK solver node? Or other 
> nodes to lighten the burden on content developers? Joe mentioned 
> something about assigning joint weights automatically. Is that a 
> node, or option in node?

Sure, you look out there and see the IK in the tool, which really 
helpsif you are making ketframes, or tyrying to do a realistic 
simulation. You push a button and the thing gets rigged. Fantastic.

Ihope you will see that all the basic stuff needed to fully document 
those items of interest are in the or can be added to, the HAnim.

> ...
> I'm not sure I understand the difference between maya and blender 
> when it comes to root nodes. Isn't that something an exporter could 
> add, if missing?

Look at LOA0. All there is is root node. The root node navigates the 
position and orietation of the skeleton-space, within the Humanoid 
space.

> ...
> I might have time in 2017 to do blender exporter for hanim.

DOes freewrl do hanim? Can you do the skinnng?

> I saved the old campbell barton python x3d exporter, which I hacked 
> several ways before. Although it seems like a struggle to keep it 
> working in latest revisions of blender: plugin api was a moving 
> target for a while, didn't work last time I tried it:
> https://sourceforge.net/p/freewrl/git/ci/develop/tree/freex3d/blender_scripts/
>

See what you can get and we can convert to X3D hanim.
Warning, getting the skin bindings may be hard.
All the joint (or bone) vertex vindings and weights may be locked up 
in a script (as I saw in Unity).
In general, the bindings are the most carfully kept secret of the 
avatar.

>

Thanks,
Joe
>
>
> ________________________________________
> From: x3d-public <x3d-public-bounces at web3d.org> on behalf of Joe D 
> Williams <joedwil at earthlink.net>
> Sent: November 11, 2016 7:20 AM
> To: Roy Walmsley
> Cc: 'Humanoid Animation (H-Anim) Working Group'; 'X3D Graphics 
> public mailing list'
> Subject: Re: [x3d-public] [h-anim] Basic Animataion data
>
> Thanks, Roy
> Repository at:
>
> http://www.graphicsgems.org/
>
>
> ----- Original Message -----
> From: "Roy Walmsley" <roy.walmsley at ntlworld.com>
> To: "'Joe D Williams'" <joedwil at earthlink.net>
> Cc: "'X3D Graphics public mailing list'" <x3d-public at web3d.org>;
> "'Humanoid Animation (H-Anim) Working Group'" <h-anim at web3d.org>
> Sent: Friday, November 11, 2016 3:41 AM
> Subject: RE: [h-anim] Basic Animataion data
>
>
>> Hi Joe,
>>
>>
>>
>> Great summary and references. However, one of the references you
>> listed is
>> incorrect. You wrote:
>>
>>
>>
>> Schlag, 1994] John Schlag. Using geometric constructions to
>> interpolate
>> orientation with quaternions. Graphics Gems IV, pages 230-236, 
>> 1994.
>>
>>
>>
>> In fact, the article by John Schlag you are referring to is:
>>
>> John Schlag. "Using Geometric Constructions to Interpolate
>> Orientation with
>> Quaternions". In Graphics Gems II, Academic Press, 1991, pp.
>> 377-380.
>>
>>
>>
>> Consulting Graphic Gems IV, pages 230-236, there is an article
>> "Fiber Bundle
>> Twist Reduction", by Ken Shoemake at that location.
>>
>>
>>
>> Regards,
>>
>>
>>
>> Roy
>>
>>
>>
>>
>>
>> -----Original Message-----
>> From: h-anim [mailto:h-anim-bounces at web3d.org] On Behalf Of Joe D
>> Williams
>> Sent: 11 November 2016 01:30
>> To: X3D Graphics public mailing list <x3d-public at web3d.org>;
>> Humanoid
>> Animation (H-Anim) Working Group <h-anim at web3d.org>
>> Subject: [h-anim] Basic Animataion data
>>
>>
>>
>>
>>
>> glossary:
>>
>> Transform rotation value
>>
>>
>>
>> * Euler angle - typical mocap-derived
>>
>>  animation data = x,y,z
>>
>>     (x,y,z = degrees)
>>
>> Mocap data represents:
>>
>>  * Bone or Segment orientation
>>
>>      or
>>
>> * parent Joint rotation
>>
>> Either interpretation represents
>>
>> same animation
>>
>>
>>
>>
>>
>> * matrix = axis-angle = x,y,z,a
>>
>>     x,y,z = radians,
>>
>>     a = scale factor
>>
>>
>>
>> X3D Default
>>
>> Easy to hand-edit
>>
>>
>>
>>
>>
>> * Unit quaternion = w,x,y,z
>>
>>
>>
>>      w+x+y+z = 1 = unit q
>>
>>
>>
>> Most common authoring tool internal and transport form Easy to add
>> unit
>> quats for realtime interpolated animation Can use hand-edit as
>> w=scale
>> factor and x,y,z are axes  if careful to maintain uq
>>
>>
>>
>>
>>
>> * Standards-Track Lossless conversion between forms.
>>
>>
>>
>> First, please notice that this Animation data is aimed at 
>> describing
>> keyframe animation of a set of hierarchal "Joint" nodes.
>>
>>
>>
>> Animation data is provided as arrays of:
>>
>> keyframe root position data,
>>
>> keyframe rotation data for each joint,
>>
>> number of keyframes,
>>
>> target key time frame interval
>>
>>
>>
>> The data is used by defining the root position and each joint
>> rotation at
>> each key time. For film and video there is often no need for a list
>> of key
>> times indexed to each keyvalue because the frame interval is fixed.
>>
>>
>>
>> Notice that typically, there is no consideration for non-monotonic
>> key
>> times, although it is highly unlikely to encounter an X3D browser
>> that does
>> not allow non-monotonic key times. Why, allow non-monotonic key
>> times?
>> Because it is usually convenient for the author to consider each
>> joint
>> individually and individually define the interval between 
>> associated
>> keyvalue data points, and to minimize the actual number of key 
>> times
>> and
>> thus keyvalues.
>>
>>
>>
>> At this time, in order for an X3D HAnim to use most types of mocap
>> data it
>> would only be required:
>>
>>
>>
>> 1. Connect the joint names
>>
>> 2. Convert the input data form to axis-angle, 3, Construct the
>> interpolators
>> and timers for each joint
>>
>>
>>
>> This style of animation depends mainly or solely on application of
>> individual rotations to individual joints in a set of hierarchal
>> 'Joint'
>> nodes at each key time to produce the keyframe. Current best
>> practice for
>> this style of animation uses unit quaternions due to efficiency.
>> Even though
>> the most throughly complete technical solution remains axis-angle
>> matrix
>> transformations, the unit quats are easy to transport and used
>> natively in
>> many animation creation and development tools.
>>
>>
>>
>> So, the choice of gltf to only transport quaternions should cause
>> X3D and
>> HAnim to address the idea of built-in quaternion support. As is
>> well-shown,
>> the maths of these three forms are well defined and conversion
>> interactions
>> are well understood.
>>
>>
>>
>> * At this time we are creating prototype support for BVH x,y,z ffps
>> import,
>> it seems only reasonable to encourage import of both Euler angle 
>> and
>> unit
>> quaternion forms.
>>
>>
>>
>> Background
>>
>>
>>
>> Quaternions, Interpolation and Animation Erik B. Dam Martin Koch
>> Martin
>> Lillholm  <mailto:erikdam at diku.dk> erikdam at diku.dk
>> <mailto:myth at diku.dk>
>> myth at diku.dk  <mailto:grumse at diku.dk> grumse at diku.dk Technical
>> Report
>> DIKU-TR-98/5 Department of Computer Science University of 
>> Copenhagen
>> Universitetsparken 1
>>
>> DK-2100 Kbh Denmark
>>
>> July 17, 1998
>>
>>
>>
>> Interesting Material.
>>
>> I have a copy of a pdf.
>>
>>
>>
>> Appendix A - Conventions
>>
>>
>>
>> Coordinate system
>>
>> (slightly reworded by me to aim at X3D and HAnim)
>>
>>
>>
>> X3D and HAnim use a right-handed coordinate system. In computer
>> graphics
>> some use a left-handed coordinate system. This allows the z-axis to
>> point
>> "into" the screen which seems natural for some styles of authoring.
>> However, since X3D primarily uses coordinates for animations driven
>> by
>> mathematical derivations and for Humanoid animations, VRML and X3D
>> have
>> chosen to use the mathematical standard, the right-handed 
>> coordinate
>> system.
>>
>>
>>
>> Rotation.
>>
>>
>>
>> This means that the default Humanoid pose gaze is facing +z.
>>
>> The +z axis for the character points "out" of the screen,
>>
>> +y is up,
>>
>> and +x is towards the character's left, viewer's right.
>>
>> The default X3D viewpoint is oriented to look toward -z.
>>
>>
>>
>> The direction of positive, or increasing rotation about an axis is
>> obtained
>> by the right-hand rule:
>>
>>
>>
>> Hold the axis with right hand and the thumb pointing in the 
>> positive
>> direction of the axis.
>>
>>
>>
>> Adding positive rotation will rotate in the direction of the
>> fingers.
>>
>>
>>
>> It just happens that HAnim Joint rotations can be vizualized in
>> terms of
>> x=pitch, y=yaw, and z=roll.
>>
>>
>>
>> Again, hold the Joint, as if you were holding the appropriate axis
>> with
>> right hand and the thumb pointing in the positive direction of the
>> axis.
>>
>>
>>
>> A positive rotation will rotate the Joint and move child 
>> hierarchies
>> in the
>> direction your fingers are wrapped around the axis.
>>
>>
>>
>> For instance, consider a roll (positive z-axis) animation of a
>> Joint, as
>> would be appropriate for HAnim right hand fingers from the default
>> pose..
>> The fingers are pointing down and thumb pointing toward +z. To move
>> the
>> fingers as if grasping the z-axis, rotate the finger joint(s) to
>> increase
>> their z-axis rotation.
>>
>>
>>
>> Amazingly enough, this results in the character's right hand
>> wrapping around
>> the z-axis with the thumb pointed +z, and illustrates the 
>> right-hand
>> rule
>> perfectly.
>>
>>
>>
>> Incidentally, this means that to get proper orientation for a gaze
>> from the
>> 'standard' eyeball location, then you can yaw the viewpoint
>>
>> +pi radians.
>>
>>
>>
>> Transport Animations
>>
>>
>>
>> In order to transport animations between 'standard' H-Anim
>> characters,
>>
>> to accomplish the HAnim 'standard' initial pose prior to animation
>> the
>>
>> Joint is defined to be at X3D default orientation, +z is out of the
>>
>> screen, +y is up, and +x is towards the character's left.
>>
>>
>>
>> The default initial value for all Joints:
>>
>> X3D axis-angle (matrix) form is 0 0 1 0
>>
>> and for unit quatenions is 1 0 0 0
>>
>>
>>
>> An illustration:
>>
>> [standard X3D default transformations
>>
>> X3D coordinate axes]
>>
>> z (+ out toward viewer)
>>
>> y (+ up)
>>
>> x (+ character's left)
>>
>> Rotation about z brings x into y
>>
>> Rotation about y brings z into x
>>
>> Rotation about x brings y into z
>>
>>
>>
>>
>>
>> Schlag, 1994] John Schlag.
>>
>> Using geometric constructions to
>>
>> interpolate orientation with quaternions.
>>
>> Graphics Gems IV, pages 230-236, 1994.
>>
>>
>>
>> [Shoemake & Du
>>
>> , 1994] Ken Shoemake & Tom Du
>>
>> . Matrix animation and
>>
>> polar decomposition.
>>
>> <ftp://ftp.cis.upenn.edu/pub/graphics/shoemake/polar-decomp.ps.Z>
>> ftp://ftp.cis.upenn.edu/pub/graphics/shoemake/polar-decomp.ps.Z,
>> 1994.
>>
>>
>>
>> Shoemake, 1994b] Ken Shoemake. Quaternions.
>>
>> <ftp://ftp.cis.upenn.edu/pub/graphics/->
>> ftp://ftp.cis.upenn.edu/pub/graphics/-
>>
>> shoemake/quatut.ps.Z, 1994.
>>
>>
>>
>> Thanks and Best,
>>
>> Joe
>>
>>
>>
>>
>>
>> _______________________________________________
>>
>> h-anim mailing list
>>
>> <mailto:h-anim at web3d.org> h-anim at web3d.org
>>
>> <http://web3d.org/mailman/listinfo/h-anim_web3d.org>
>> http://web3d.org/mailman/listinfo/h-anim_web3d.org
>>
>>
>
>
> _______________________________________________
> x3d-public mailing list
> x3d-public at web3d.org
> http://web3d.org/mailman/listinfo/x3d-public_web3d.org
>
> _______________________________________________
> x3d-public mailing list
> x3d-public at web3d.org
> http://web3d.org/mailman/listinfo/x3d-public_web3d.org