Name : Marcella Tanzil
Email: tanzil.10@osu.edu
Offices:
OSU : 413 Dreese Lab
CMU: 4228 Newell-Simon
Phone no : (614)316-3552
Please click on the icon below to send me an email:
|
Week ten (Aug
17th - Aug 23rd)
For this final week, I am pretty much tweaking the backflip guy for
the next experiment. I also set up the blobs to
have better background so that the viewers would have
more cues with respect to the size of the
blobs. We are not sure whether we will be using
the second version of the blobs for the survey
depending on the result of the results of the
first survey. The backflip, however, would be
used next after the first blob experiment.
Week nine
(Aug 10th - Aug 16th)
We ( my
teammate and I ) are preparing for our final presentation.
We are also
assigned some tasks as follow:
1)
run the
current experiment. They've applied for IRB approval
on that but as I think
you saw in email that will take a bit
of time. Did you
both do the NIH test? If so, you need to
get those certificates
to the person who applied for the IRB
approval or it will get
bounced.
2) run the same blob
motion but now with the chair and palm
tree in the scene (both
in the still and in the movies) and
see if that changes the
response. To do that, you need to
get the chair and palm
tree in scale to each other. We should
use "real" units
(meters) and look up the size of a palm tree
and a chair of that
style and really get this right, not just
eyeball it.
Google should get you that info quickly enough.
Maybe scatter a few palm
trees of slightly different sizes
in the background
(rotating them so that they don't all look
identical). I
think that the palm tree was much too short because
it was in the foreground
and you wanted to be able to see the top
of it.
3) do the same thing but
with some dramatic motion for an animated
person driven by mocap
data. I'm not sure that we ever converged
on what that motion
might be. I was hoping for a flip -- did you
get that motion cleaned
up? Failing that, I would try a broad
jump. I want
something with a long flight phase so that people
will have a better
chance of seeing that the motion is scaled incorrectly.
If the character is
always in contact with the ground (as in the
gorilla boy), it would
be physically possible to be applying forces
that would make the
character move faster (or slower) so it isn't
so clear that the motion
is scaled incorrectly.
Hopefully, with these,
we can see:
1) scaling correctly
matters (hopefully supported by all three studies)
2) adding objects for
scale makes it a stronger effect
3) it is a stronger
effect for human motion than for non-humanoid
motion.
We were done with task
one and starting on task two for this week. We also did our
presentation on Tuesday ( Aug 16th). It went well.
Week seven
and week eight (July 27th- Aug 9th)
July 27th - July 28th
We finished everything
and make sure
everything is okay with the blob movies. We also modified some of the
stuff according to Sara (another faculty that collaborates with Prof.
Hodgins). We want to make sure everything is good to go with the survey
before we left for SIGGRAPH
July 29th -
Aug 6th
I went to the
SIGGRAPH 2005 Conference at Los
Angeles Convention Center. I really enjoy it and would love to go there
again next year.
It gives me a lot of new knowledge as well as ideas that I have never
thought about it before.
The part that I enjoyed most from the SIGGRAPH is the fact that they
have courses which was the most interesting part.
Week six
(July 20th -
July 26th)
We finalized
and tweaked the blob movies so that they are presentable. We ended up
with 9 blob movies as follow:
big size blob
with slow motion (the right scale)
big size blob
with the medium motion (wrong scale)
big size blob
with the fast motion (wrong scale)
medium size
blob with slow motion (wrong scale)
medium size blob with the medium motion (right scale)
medium size blob with the fast motion (wrong scale)
small size
blob with slow motion (wrong scale)
small size blob with the medium motion (wrong scale)
small size blob with the fast motion (right scale)
We also
created a survey for the blob movies and we will have quite a few
people to participate in the survey.
Vasu ( my teammate) and I attended the meeting to discuss the design of
the survey so that the survey would give the accurate results.
Week
Five (July 13th - July 19th)
We finished
up the the "Afraid" and "Confidence" movies to be sent to University of
Virginia so that they can decide whether they want to go further with
the research.
We rendered all the movies ( total there are around 5-6 of them) with
each of them having different scale and speed. We did this so that it
would give us more information of what would happen when we do not have
the right scale of gravity with respect to the mass of the object/
Week
Four (July 6th - July
12th)
This week, we
started to
make movies that are going to be used for the experiment to see whether
different sizes and wrong motions (not scaled properly) would distort
the original motions.
We tried it
using
different models in order to demostrate this. We basically have
blobs and also a human figure that has different kind of motions. These
motions are gathered from the CMU motion capture database. In this
case, we are using the gorilla motion and also afraid and confidence
motions.
After we
rendered all these samples into movies, we figure out
that the "Afraid" and "Confidence" motions did not show very
significant difference between the different sizes of the body. Hence,
we dont use these models any further.
Week
Three
(June 29th - July 5th)
In this week,
Prof. Hodgins gave
us few papers to
read about the dot patterns on human motions etc. The titles are as
follow:
Proffitt,
D.R., Bertenthal, B.I., & Roberts, R.J., Jr.
(1984). The role
of occlusion
in reducing multistability in moving point light
displays.
Perception
& Psychophysics, 4, 315 323.
Bertamini, M.
& Proffitt, D.R. (2000). Hierarchical motion
organization
in random dot
configurations. Journal of Experimental
Psychology: Human Perception
and Performance, 26, 1371-1386.
Proffitt, D.R.
& Bertenthal, B.I. (1988). Recovering
connectivity from
moving point
light displays. In W.N. Martin & J.K. Aggarwal
(Eds.),
Motion
understanding: Robot and human vision, Hingham MA:
Kluwer.
I also have a
chance to learn the process of motion capturing, starting
from capturing the actor, cleaning up the data, converting the data
into maya binary file that has the same motions and render it to
movies. I have some of my movies posted in my other website :
http://www.cs.cmu.edu/~tanzil.
My teammate
and I also did some research on occlusion because this
technique is needed for making the dots more recoqnizable.
Week
Two (June 22nd - June
28th)
We continued
working on the
third task. We wrote few
different MEL scripts in order to do this. I'm familiar with MEL
scripting, but not that much that I coulf figure out everything at
once. I read some tutorials and also tried to figure out the syntaxes
for different things.
I face a
problem that the coordinates of the dots are not world
coordinates, so when I used that coordinates to create new dots, it
gave me problems. We ended up using the locators in every dots to be
able to get the world coordinate of each dot. It worked.
Week One
(June 14th - June
21st)
My teammate,
Vasu and I were given a strter
project by Prof. Jessica
Hodgins. It's a project that deals with Maya Software, a software used
for computer animation. We basically have 3 tasks that we have to do
with the files that were already provided for us, which is basically
dot patterns files from Motion Capture Lab.
The three
tasks are as follows:
- We have to change the background colors and
delete ( hide) the walls and the floors so that all the dots can be
seen clearly with just grey background.
- We have to invert the figures 180 degrees so
that it would be upside down.
- The third one, which we spent quite a lot of
time workin on this is to provide the scrambled motions of the dots in
which they still have the similar trajectory motions. It means that The
dot initial position would be randomized, for example : the dot that
represented elbow earlier would have the initial position swapped with
the head's dot, so the movement of the head would follows the elbow's
trajectory and the elbow's movement would follow the head's trajectory.
For this week,
we finished the first 2 tasks but not the last tasks
even though we already started on it.
|
