Design 950 and 993 Summer 2004
This quarter I worked
on creating the mocap movie files that fit within the empty movie clip symbols
in the Canine
Lameness Learning Module. (Follow the link to view all the finished mocap
In sum, this entailed:
- Importing the .fbx
cleaned mocap files into Maya. There is one Maya file for each take.
- Creating a skeleton
to fit within the marker data in each take.
- Freezing Transformations
on the skeleton after aligned with mocap markers.
- Adjusting the local
rotation axes on every bone within the skeleton to have Z pointing down the
bones and either all X axes or all Y axes pointing in one direction. Had to
grab each LRA and rotate individually.
- Creating Polygon
cylinders and boxes to fit onto the skeleton.
- Freezing transformations
on the geometry.
- Selecting each piece
of geometry and its corresponding bone and Rigid binding the geometry to the
- Adding RPik handles
to every bone in the skeleton except the root and tail bones.
- Point constraining
each ik handle to its corresponding mocap marker/s at each joint.
- Point constraining
the root to the sacrum mocap marker.
- Adding spline ik
handles to the tail bones.
- Creating cluster
handles on the Control Vertices of the spline IK's curve.
- Point constraining
each cluster handle to a mocap marker.
- Playing the animation
to see if I missed anything.
- If the take had stair
case markers associated with it, I created polygon faces to fit within the
mocap markers to make up all the sides.
- I merged the edges
of all the poly faces to make the stair case one piece of geometry.
- I made up shaders
for the geometry. Note: I originally put one red shader on both right legs,
and a cyan shader on both left legs of each dog, and rendered the takes all
the way through to .swf files. However, it was hard to distinguish front legs
from back legs in the posterior and frontal view movies. So, I changed the
shaders in each previously rendered Maya file to have a slightly lighter cyan
on the front left leg, and a pink or light red on the front legs. I then,
re-rendered each Maya file and took it through the process explained below.
This consumed extra time, but I think it was well worth it.
- I selected each piece
of geometry and assigned a shader to it.
- I created 5 directional
lights around each dog: one overhead, one on the left side, the right side,
the back, and the front.
- I parented these
lights to the T12 thoracic marker.
- I created 4 cameras
around each dog: one overhead, one on the right side, the back, and the front.
- I parented each camera
to the T12 thoracic marker.
- If the dog did not
travel in a straight line, I had to key frame the camera movement to keep
the dog in the right perspective.
- On most of the takes,
the length of the bones were too long in some poses and the joints would flip
out of place (at unnatural angles). This happened because the suit stretched
and contracted while the dog was moving. The markers didn't always stay over
the joint. In the Maya mocap files, the leg bones are being controlled by
the stretching and expanding marker data, though the length of a dog's limb
doesn't change. I had to pick one pose (one frame) to rig each dog on. I experimented
with different rigging techniques for about a week and a half, then finally
tried baking the keys on all the ik handles. I was then able to delete the
bad keys, grab individual ik handles, and key them into a position I liked.
This was somewhat time consuming. See
example .mov of flipping joints (5MB file). Look at the front feet for
the flips. The pink foot flips at the top of the stairs as well as the light
blue one. The light blue one also flops after the dog has crossed the stairs
and turned around.
- When I was ready
to render out of Maya, I adjusted each light so that the one shining on the
side of the dog that was being rendered was the brightest (e.g. 1.2 intensity).
The other lights were taken down to about a .4 intensity value.
- The .tga still image
files that were rendered from Maya for each side, were imported into Premiere.
From here, I could add transitions, repeat takes, and manipulate the speed
of the take (from 1x to .25x).
- I exported the Premiere
files as uncompressed .mov files. 640 x480. I ended up going with 30 fps instead
of 15fps. My movies looked like they were playing too fast at 15 fps.
- Each movie had to
be taken into Sorenson Squeeze to compress it, resize it, and reformat it
for Flash. They were converted to 480 x 360px .flv files.
- I created individual
Flash movies (.swf files) by importing the .flv flash video files into .fla
files. I reused the Flash movies by simply saving them under a different name,
deleting the existing .flv and importing a new one. I then added frames if
needed and published as .swf.
- The .swf's were
placed into the proper directory structure on the web in order to work with
the main canine learning module movie. As long as I followed a strict naming
convention set up last quarter, I had no problems.
- I added some text
to the main Canine Learning Module to describe Hip Dysplasia. I used the following
book as a reference: Diagnosing Lameness in Dogs by Leo Brunnberg, 2001.
- I visited with Jon
Dyce every 2-3 weeks to show him the work. He would give me information on
which leg was lame and what he was looking for. Once all the files are in,
he will sit down with me for an extended period and refine the text part.
- I backed up files
just about every night.
Currently, I am keyframing
ik handles on the last of the mocap takes (CCLR- Stairs) in Maya. I plan to
finish this and have it rendered by the end of the last week in August. During
the first 3 weeks of September, I will get the video captured, rendered, and
put into the flash learning module as well as all the text. My goal is to have
the learning module completed by the beginning of Autumn Quarter. I then need
to work on creating the survey. I plan on creating one mocap file with muscles
contracting to include in the survey. This will be done in Maya. I'll use the
normal trot take that already has the 3-D skeleton bound to it. See
the 3-D skeleton normal trot movie in its current state (1MB .swf file embedded
in html - Quarter Speed). Download
just the .swf file.
I'd also like to have
a quiz for the students with "unknown" mocap data that they haven't
seen before. I will ask them which leg is lame to test the effectiveness of
the motion capture. If I put unknowns in, that means I have to finish the second
stage of cleaning on those motion capture files. More time consuming work within
Motion Builder. After cleaning them, I'll have to bring them into Maya and rig
and render them like I've been doing this quarter. Then create the Quiz in Web
CT or as form mail on my internet site.
I've still got a lot
of work to do.