G5-Update3
The Input
- We shot new digital stills of stadium to create a model of the football field
We were initially going to use a wide-angle camera to get shots of the football stadium, but we realized that it didn't matter if we used a regular or wide angle lens, it would just result in more or less shots taken with the same information there.
We will later have to go back and re-film the field because it currently does not have hash marks or numbers for the yard lines.
We will also be doing more view points later to get a view from every side of the stadium.
- Play description for modeling
- Camera calibration data for compositing
The Effects
- Computer model of a football field with computer generated players
- Modeled players executing various football plays from the GT play book
- (possible) composited/juxtaposed against actual football data
The Layers
- The computer generated model of the football field
- The computer generated players
- (possible)backgrounds or stadium
- (possible) fans?
Progress that has been made
We attempted to do some calibration of the screen caps taken from the video clips of the different
angles. We found that it was very hard to accurately use any of the markings on the field, due
to a combination of poor lighting and poor field markings. Also, we were only able to use the markings
between the 45 yard lines so we had a very limited selection of markings on the field. Work has been
started on rendering out a single play using cylinder men and MEL. We still need to get the model of
the field from group 11
Additionally, we have created a basic visually representation of the field and overlaid that onto a
model in Maya so that we can use Mel scripting to create and move the players around on the field to
simulate a play from the play book. Here is a screen cap of the current working model. The current
model was taken from google earth imagery. We plan to create a more realistic model at a later date.
The play script generator is nearly finished. A slight problem we encountered was simply figuring
out all the commands and shortcuts to use in the generation. We also switched from C to Java as the
language that is used to create out play scripts.
REGULAR(PRO) FORM
Problems we encountered
We encountered all kinds of problems when trying to use MATLAB to calibrate our camera. Poor lighting
during the shoot, poor markings on the field, and a limited selection of markings that were present in
all shots contributed to us scrapping our initial footage.
Transforming the panorama to fit exactly with the model has been another problem we are trying to
overcome. Currently we are simply using a set of google earth pictures pasted together and overlayed
on the model in Maya. Below is a screen cap of how the panorama was looking. As you can see the yard
lines do not match up at the top or bottom, depending on how we transformed the picture.
Panorama not lining up with the Maya model:
Curvature in the panorama of the stadium, incomplete footage:
(These panoramas have been trashed in favor of more complete ones in better lighting)
Problems we may encounter
Rendering the players through Mel scripting could be less flexible than we initially wanted. Depending
on the time available, it would be great to export the scene to openGL and have some more flexibility in
rendering the play.
Timeline
04/12 – Preview Showing of the Final Project
04/17 – 4th and Final Update Due
04/26 – Presentations
Links to this Page
- DVFX 2007 Groups last edited on 2 May 2007 at 12:51 am by coppola-win.cc.gatech.edu
- G5-Final Report last edited on 3 May 2007 at 1:15 pm by c-24-99-1-204.hsd1.ga.comcast.net