******************************************************************** * * * A System for View-Dependent Animation * * Parag Chaudhuri, Prem Kalra and Subhashis Banerjee * * * ******************************************************************** The four movie clips are numbered as they are presented in the paper. We recommend that the clips are viewed in the order they are numbered after reading the corresponding discussion in the paper and in this file. ------------------------------------------------------------------------------ Clip 1: These image shows "all" the keyframes for clip_01.avi. It contains all the sketched poses and the corresponding rendered frames. In each of the sketches Hugo is posed and deformed differently. While some of these deformations can be obtained directly by reconstructing the skeleton pose, at other places the mesh needs to be deformed to match the sketched character more accurately. The animation sequence has the camera going around Hugo cycling through each of the key viewpoints with brief pauses at the top and bottom key viewpoints to highlight the correctness of the full projective camera recovery. Hugo deforms accordingly in response to the camera. When the camera is to the right of Hugo, his right leg undergoes a marked mesh deformation which cannot just be obtained by skeletal pose recovery. This is as per the corresponding sketched view. Other poses have less amount of mesh deformations. ------------------------------------------------------------------------------ Clip 2: The image shows "only" sketched pose used in the "view-dependent part" of clip_02.avi. Along side the sketch is the rendered keyframe shown from a slightly different camera angle. This clip starts with a pure IK sequence and then the camera begins to move to indicate the start of the view-dependent part of the animation. At this point in time Hugo is suspended in mid air doing a back flip. The camera cycles through three key-viewpoints, one of which was obtained from a sketch, while the other two were the last and first poses of the beginning and ending IK sequences. The camera reconstructed is a full projective camera, and the sequence also has mesh deformation as can be seen in the curved limbs of Hugo. The sequence ends with another IK sequence. ------------------------------------------------------------------------------ Clip 3: The image shows "all" the keyframes for the "view-dependent part" of clip_03.avi. Note that the sketched poses in these are "mannequin" sketches. At the side of each sketch is the corresponding rendered keyframe. We stress here that all the changes in the character's pose is in response to the camera movement and it should not be confused with a rigid rotation of the character. ------------------------------------------------------------------------------ Clip 4: The image contains "some" of the keyframes for the "view-dependent sequences" in clip_04.avi. The first two sketched poses correspond to the "ballet jump" view-dependent sequence, while the later two sketched poses correspond to the ``pirouette" view-dependent sequence. At the side of each sketch is the corresponding rendered keyframe. This clip has three view-dependent sequences. The first has the hand trying to do a jump very characteristic to a ballet sequence. Notice that the legs (of the character) do not stretch as far so as to execute a perfect ballet jump the first time, so the character retries the jump. This second attempt of the character has been generated by changing the camera angle and replaying the same sequence again. It can be clearly seen that the legs stretch further apart this time, and hence we clearly demonstrate deformation changing in response to a changing camera. As a final demonstration of the elegance of the view-dependent animation technique the character executes a "pirouette" (a spin move in ballet) and it can be seen as the camera also rotates independently of the character, the character clearly deforms "while rotating", in response to the camera movement. The sketches were made keeping in mind only the movement of the camera around the character. The rotation of the character itself was added during the animation. The deformation of the character is only due the camera movement around the character. The rate of rotation of the camera and that of the character itself are different and hence are not ambiguous during the animation. ============================================================================== NOTE: All the movie clips are encoded using the standard DivX Codec. The codec can be obtained from http://www.divx.com. The clips can be played using Mplayer on Linux or Windows Media Player on Windows or any other player which understands the DivX format. All the images are standard JPEG images. ==============================================================================