Joshua Glazer Kenneth McCracken Alex Sverdlov [email protected] [email protected] [email protected] vs. Machine6.837 -- Introduction to Computer Graphics Final Project -- team 01 November 30, 1999 Abstract This paper presents an account of experience gained by implementing a final project for Course 6.837Introduction to Computer Graphics at MIT. The team consisting of two seniors and a graduate studentcreated a video of a live person performing martial art techniques against a computer generatedadversary. The film was shot using a Canon XL-1 digital movie camera, and subsequently edited usingPoser 4.0 and Final Cut Pro 1.0. These and other tools were used to render the digital character in thecontext of filmed scenes, and to fine tune the interactions between the two opponents. It was the group’sexperience that the significant portion of technical challenges were eventually overcome through skillfuluse of the tools. However, a few of the project’s objectives needed to be met by writing and using Croutines to manipulate the decompressed screen shots directly. The finished video was accompanied bya synchronized soundtrack composed by Joshua Glazer, and concluded with a few of the more humorousout-takes from the filming. Keywords: computer graphics, 6.837, MIT, video, Poser, Final Cut, Tsai method, martial art, modeling,compositing, c code Introduction The original idea for this project was conceived by Josh Glazer, with Ken and Alex signing onto theproject during the team forming period. It was the consensus among the team members that creating avideo of this kind would put to use skills gained in the course teach us a thing or two about the tools pertinent in the computer graphics industry be rewarding from a creative standpoint Indeed, in the process of making the film we had to draw extensively upon 6.837 course material todetermine lighting, shadowing effects, plan camera and lamp positions, as well as to accomplish correctcompositing between the digitized character and the filmed scenes. We have also gained experience ofresearching industry publications that dealt with solving problems similar to our own -- most notablypapers by Roger Tsai on camera calibration [5], and on analysis of 3-D time varying scene [6]. We havelearned to use the results presented in these articles as an aid in understanding and meeting our project’schallenges. In turn, we hope that the present report of our efforts and mistakes may serve as a point ofreference for others in the future. Goals Our final goal is a 1.5 - 2 minute digital video depicting a martial arts contest between twoopponents. While one of the fighters is human, the other is a modeled animation superimposed onto thescene. The key difficulty of this goal was making the combat appear realistic and engaging,encouraging the suspension of disbelief. The scenes had to be modeled carefully, with attention paid toissues both obvious and subtle. While much time was spent getting such macro effects as shadows tolook convincing, many smaller details of animation were also implemented. For instance, Fred’s (hehad to get a name, since we couldn’t really say ’computer-generated-animated-3D-character’ hundredsof times on end) mouth moves synchronously with his speech. He also shifts his eyebrows, and blinks atrandom intervals, adding to the realism of the scene. To accomplish the necessary effects, the main goalwas partitioned into the following tasks: storyboarding the battle before hand shooting the film itself importing and editing the footage modeling Fred’s anatomy, his clothes and weapons teaching Fred’s body to react realistically to the techniques dealt out by Sensei Godfrey Inniss(human character) modeling the interactions of the fighters’ weapons and body parts during battle (compositing of thepixels) tracking the position of the camera compositing Fred with the live video creating the shadowing effects of the animated figure in the live scene and finally adding an award winning soundtrack Each of these items was addressed in turn, with the next section of the paper detailing our successes andfailures at each of the stages. AchievementsStoryboarding Our first material step on the road to this blockbuster was creating a storyboard. We sketched thescenes that introduce the characters, planned the positions and angles of the shots depicting the battlefrom various points of view, and scripted a quality dialog in the proud tradition of imported kung-fumovies (see appendix H). As a location for the shooting, we chose the seldom trespassed corridor in thebasement of Building 9 at MIT. This hallway, with its chain-linked fences and extensive pipingprovided an original and engaging setting for the conflict. Our plan was to shoot the live video and to insert the animated character into the backdrop later. Using a stand-in adversary to temporarily cover for Fred would require the eventual removal of onefighter from the video. This approach would risk possible loss of data as per the ray tracing [1]paradigm, in scenes where Fred did not completely cover his stand-in. Consequently, we opted to havethe live character fight a nonexistent foe. Striving for realism, we sought out a consultant (soon turnedstar) Sensei Godfrey Inniss. He helped us complete the storyboard with specific techniques drawn fromhis background in Jui-Jitsu. With the plot in place, we decided on the finale, and scheduled the shoot. Filming and Editing Each scene in the movie was shot twice. First, with Godfrey practicing a move on one of us acting asan ’ookie’ (a recipient of technique), and then executing the same strike or throw against an imaginaryopponent. The first version served as a template for animating Fred’s movements, while the second takewas the actual footage used in the finished film. While shooting, we were challenged to try to limit the number of sequences in which the camera wasmoving, while striving for the best possible coverage of the action. This was a concern because weknew that modeling the details of the scene shot by a moving camera was significantly more timeconsuming. However, many sequences in which Fred appears alone did involve motion of the cameraitself in order to give the viewer a more exciting perspective. To provide light for the filming, we used two bright 300W work lamps. We tried to keep thesestationary to simplify modeling. Still,
View Full Document
Unlocking...