I had read a while ago that you will need to treat each matchmove differently, and i am starting to see that this is true. I aslo read that there are two types of matchmover, the logical and the artistic. The logical matchmover will use all available information to solve a shot, and if no information is present, make educated guesses about length and distances ect. then there are the artistic kind that will engineer solutions to solve problems.
I have just finished my third matchmove and it has been fraught with problems. Firstly the shot was very shakey, the second, there were very few good tracks. After trying and giving up on an easy auto track solution, I tried a few well placed manual tracks. this held no solution to anything as i could not track forward or back for more than a frame or 2. After literally spending a day on this, I thought do or die. I manually tracked every keyframe for every track on this shot. here is the proof. And let me tell you, it was the only way!
So after a long time doing this i finally got a nicely solved co-ordinate system working in matchmover. Now I had the usual task of balancing the focal length of the shot with the real measured geometry and blocking it all in. I had another problem here. Due to the huge lack of trackable information, or due to the poorness of the tracked points, the deapth would not sync with the real geo 100%. and by 100% i mean, one plane of geometry.
everything was messured fine, and no ammount of unrealistic edits to the geo would solve this problem for every frame. Finally i resorted to animating a planar surface to hold this troublesome edge. I have predicted that this will solve our problem however time will tell, and i have no problem re-tackeling this problem if it turns out i have not solved this problem yet.
But this shot has now been packaged and is ready to be taken by the animators! on to shot 4!!!
I have just finished my third matchmove and it has been fraught with problems. Firstly the shot was very shakey, the second, there were very few good tracks. After trying and giving up on an easy auto track solution, I tried a few well placed manual tracks. this held no solution to anything as i could not track forward or back for more than a frame or 2. After literally spending a day on this, I thought do or die. I manually tracked every keyframe for every track on this shot. here is the proof. And let me tell you, it was the only way!
So after a long time doing this i finally got a nicely solved co-ordinate system working in matchmover. Now I had the usual task of balancing the focal length of the shot with the real measured geometry and blocking it all in. I had another problem here. Due to the huge lack of trackable information, or due to the poorness of the tracked points, the deapth would not sync with the real geo 100%. and by 100% i mean, one plane of geometry.
everything was messured fine, and no ammount of unrealistic edits to the geo would solve this problem for every frame. Finally i resorted to animating a planar surface to hold this troublesome edge. I have predicted that this will solve our problem however time will tell, and i have no problem re-tackeling this problem if it turns out i have not solved this problem yet.
But this shot has now been packaged and is ready to be taken by the animators! on to shot 4!!!
No comments:
Post a Comment