Gord.T
Still "Senior Member"
The uploaded example files show where I'm at now. Although I did center the label a bit. Here's whats going on today.
*** The cap:
Here's how I have to cap oriented. I centered the texture a little more.
All the of cap textures are of the same size and are placed in 3D space with the same coordinates.
*** The label:
I ran into a huge tecnical problem in this. Because there are not targets on the label the tracking program picked it's own targets using my settings. The tracking is perfect but I realized the targets making up the label were actually frontally flipped 180 degrees. As if the back of the bottle was tracked and not the front.
I have some theories about. If anyone has had a similar experience and knows a solution let me know.
So here's the kicker. It is now impossoble to use a properly oriented cylindrical label shape because of the way the tracking camera solved its movements. While the tracking prediction targets themselves are stable, the line of sight from the camera to any prediction target as it moves around will not line up with a label that has its front closer to the camera than it's back.
It could be shown graphically with vectors but you could sketch it yourself if it is hard to visualize.
Bottom line is the label will always float.
Solution:
I have to fake the shape of the label. That is why it looks fake a bit.
a)I tried camera mapping a perspectivly correct label.
b)I created a new label using the existing tartgets. I kept the most critical outter edge targets in their original coordinates so as not to float. I then brought the others, the frontal targets as close to the forward position as I could without causing drift.
Like I said, if the label was in the proper shape it would float. So it is halfway between a flat label and a can shape. To further help some 3D shadowing may be apllied to the edges.
Now this is getting to be a lot of work.
Here's a shot showing a selected target on the front of the label and how it is positioned in the opposite depth direction.
*** Color adjusting:
I'm leaving that to you as you'll be doing that in post anyways to match the rest of your video. But I'll play with the lighting a bit.
*** Morphing:
Because of the problems that have arisen I cannot do the morphing sequences due to time.
I was hoping you could do the sequences using MorphMan 4. Then create an image sequence of .bmp files.
Put those in a zip file to lower the size and upload them.I only need the transition shots (the morphman animations).
For example, You render out a 2 second morph sequence. Thats 50 frames at 25p. Put those 50 rendered pictures into a zip file and upload it. Then do the same with the next morph shot.
/I mean I'll still do the 3D part of it.
******************
update: WOHOO. I've found the perfect solution to the perspective problem. Later.
*** The cap:
Here's how I have to cap oriented. I centered the texture a little more.
All the of cap textures are of the same size and are placed in 3D space with the same coordinates.
*** The label:
I ran into a huge tecnical problem in this. Because there are not targets on the label the tracking program picked it's own targets using my settings. The tracking is perfect but I realized the targets making up the label were actually frontally flipped 180 degrees. As if the back of the bottle was tracked and not the front.
I have some theories about. If anyone has had a similar experience and knows a solution let me know.
So here's the kicker. It is now impossoble to use a properly oriented cylindrical label shape because of the way the tracking camera solved its movements. While the tracking prediction targets themselves are stable, the line of sight from the camera to any prediction target as it moves around will not line up with a label that has its front closer to the camera than it's back.
It could be shown graphically with vectors but you could sketch it yourself if it is hard to visualize.
Bottom line is the label will always float.
Solution:
I have to fake the shape of the label. That is why it looks fake a bit.
a)I tried camera mapping a perspectivly correct label.
b)I created a new label using the existing tartgets. I kept the most critical outter edge targets in their original coordinates so as not to float. I then brought the others, the frontal targets as close to the forward position as I could without causing drift.
Like I said, if the label was in the proper shape it would float. So it is halfway between a flat label and a can shape. To further help some 3D shadowing may be apllied to the edges.
Now this is getting to be a lot of work.
Here's a shot showing a selected target on the front of the label and how it is positioned in the opposite depth direction.
*** Color adjusting:
I'm leaving that to you as you'll be doing that in post anyways to match the rest of your video. But I'll play with the lighting a bit.
*** Morphing:
Because of the problems that have arisen I cannot do the morphing sequences due to time.
I was hoping you could do the sequences using MorphMan 4. Then create an image sequence of .bmp files.
Put those in a zip file to lower the size and upload them.I only need the transition shots (the morphman animations).
For example, You render out a 2 second morph sequence. Thats 50 frames at 25p. Put those 50 rendered pictures into a zip file and upload it. Then do the same with the next morph shot.
/I mean I'll still do the 3D part of it.
******************
update: WOHOO. I've found the perfect solution to the perspective problem. Later.
Last edited:
