A new version of Cineversity has been launched. This legacy site and its tutorials will remain accessible for a limited transition period

Visit the New Cineversity
   
 
Object Tracker 3D solver creates the wrong camera for my shot
Posted: 10 July 2021 02:48 PM   [ Ignore ]  
Total Posts:  6
Joined  2020-09-07

Hey there. I have a big problem. I’m trying to track these cool 3D elements to a phone I filmed. I watched a few tutorials, these:

Sean Dove Tutorial: https://www.youtube.com/watch?v=jbvVr1svCSw&t=203s
Digital Meat Tutorial: https://www.youtube.com/watch?v=uTyd378pP1I&t=1275s
Yellow Dog Tutorial: https://www.youtube.com/watch?v=QenuS_I0n-k

None of them deal w/how the camera is reconstructed for the scene with JUST the object tracker. They mention to motion track the scene first. That would be a good idea except I’m dealing with a static tripod shot filmed with a 105mm lens. There’s nothing to track that would help.

Anyway, I

- opened the motion tracker
- brought in my footage
- went to 2D tracking
- manually tracked the phone as best I could (didn’t put tracking markers on the phone - huge mistake). The track seemed alright.
- Then I brought in an object tracker, pasted my assigned tracking points, went to the ‘Reconstruction’ Tab, and hit ‘run 3D solver for object’ That seemed to work

BUT

It created a camera that was 17mm. NOT the 105mm I shot the scene with. I took off the protection tag that was on the solved camera, changed the focal length to 105, but then EVERYTHING was screwed up. My tracking points were all over the place.

I tried starting all over, setting the camera to 105mm BEFORE I brought in the object tracker. Same problem.

I don’t know what else to do! Please, PLEASE help. Somebody.

thank you


George

Profile
 
 
Posted: 10 July 2021 04:04 PM   [ Ignore ]   [ # 1 ]  
Administrator
Avatar
Total Posts:  12043
Joined  2011-03-04

Hi George,

There are limits in tracking when the shot was done without thinking about 3D tracking. I have seen many bad shots tracked with a lot of tricks, but in some clips, there is not enough information. Manual tracking the object is then the only way.

A long lens 105 mm with little movement will not do anything for you here, for the 3D main track. If the sensor was small even less so. Again, too little information so far.

I have sent you an upload link to see what the problem is. Unfortunately, I can’t tell only that it might be difficult from the image, but images don’t tell a lot.

If you can re-shoot:
Before the main part starts, it might always be helpful to move the camera to get more data. I can’t see any tracking targets on the green screen.

After the main shot, also while the camera continues in the same take, the phone could be moved and rotated slowly, so more data is available.

Place C-stands with colorful gaffer tape in the space.

Do a set survey so you know where in space the tracking targets are.

Have the measured lens data (Field of view, and the sensor data (what resolution was used). Don’t trust the printed data on the lens.

Have a distortion grid done with the same configuration, and if you use Photography instead on cine lenses, take that grid with the same distance your shot will be.

There are more things to consider. As before, I think Tim Dobbert’s book has a lot of tips; even his second edition is close to a decade old.

All the best

 Signature 

Dr. Sassi V. Sassmannshausen Ph.D.
Cinema 4D Mentor since 2004
Maxon Master Trainer, VES, DCS

Photography For C4D Artists: 200 Free Tutorials.
https://www.youtube.com/user/DrSassiLA/playlists

NEW:

NEW: Cineversity [CV4]

Profile
 
 
Posted: 10 July 2021 07:07 PM   [ Ignore ]   [ # 2 ]  
Total Posts:  6
Joined  2020-09-07

Thank you so much for responding! Really appreciate it. Excellent advice for the next shoot.

But I’m still stumped with the 3D solver. I wound up getting a really solid track but it turned my camera into a 17mm lens. I have to make it 105mm, to match the plate I shot. But when I change 17mm to 105mm, it completely messes up the track. Which makes the whole thing a waste.

What do yo think?

Profile
 
 
Posted: 10 July 2021 07:25 PM   [ Ignore ]   [ # 3 ]  
Administrator
Avatar
Total Posts:  12043
Joined  2011-03-04

Hi George,

You wrote what you already noted above. I need to see the footage and explore what you got.

Besides that, here are the training Cineversity offers:
https://www.cineversity.com/vidplaylist/motion_tracking_object_tracking_inside_cinema_4d/motion_tracking_object_tracking_inside_cinema_4d_introduction

In 2005 we had a meeting in London. A friend of mine showcased with some sticks how he would like to create a 3D tracker. The sticks symbolized the rays that finally build the triangles that the tracker uses to determine where the camera is. He developed it further and later joined Maxon. The tracker you use started a little bit before that London meeting.

I mention that, as the key is in the “rays” between the camera and the tracking points. If you move around an object in a quarter circle, these rays cross in 90º and the crossing point is super precise. If these lines or “rays” have very little change, then the point where they cross is not so exact. So, a long lens and minimal movement are the key elements to run into problems. Hence the preparation BEFORE you shoot. There is sometimes no fix in the post. The lens is evaluated as 17 while it was 105 (… again, never trust what is labeled on them, I have 50 lenses and each is calculated differently from what is labeled on them!)

I just saw the footage that was uploaded. I will check it out. I hope the lens profile chart is included.

Have a great weekend.

All the best

 Signature 

Dr. Sassi V. Sassmannshausen Ph.D.
Cinema 4D Mentor since 2004
Maxon Master Trainer, VES, DCS

Photography For C4D Artists: 200 Free Tutorials.
https://www.youtube.com/user/DrSassiLA/playlists

NEW:

NEW: Cineversity [CV4]

Profile
 
 
Posted: 10 July 2021 07:32 PM   [ Ignore ]   [ # 4 ]  
Administrator
Avatar
Total Posts:  12043
Joined  2011-03-04

P.S.: What camera was used? This is crucial!!! FPS?

There is nearly nothing in the shot that allows for anything…

 Signature 

Dr. Sassi V. Sassmannshausen Ph.D.
Cinema 4D Mentor since 2004
Maxon Master Trainer, VES, DCS

Photography For C4D Artists: 200 Free Tutorials.
https://www.youtube.com/user/DrSassiLA/playlists

NEW:

NEW: Cineversity [CV4]

Profile
 
 
Posted: 10 July 2021 09:14 PM   [ Ignore ]   [ # 5 ]  
Administrator
Avatar
Total Posts:  12043
Joined  2011-03-04

Hi George,

Even in another tracking app, I get no solution.

I do not really see how that shot can be solved. The only REMOTELY usable motion is sadly swallowed entirely in motion blur.

After 90 minutes and many different tests, I can only suggest matching it manually or reshoot.

The results are so different that I stay with the idea that the motion is too little. I assume that the camera was locked off, so the Motion tacking of the object alone would be sufficient as that moves the camera around and has the object as static. No object tracking is needed.

However, sorry, the shoot was done with zero concerns towards tracking. Nothing that can be “healed” in post.

All the best

 Signature 

Dr. Sassi V. Sassmannshausen Ph.D.
Cinema 4D Mentor since 2004
Maxon Master Trainer, VES, DCS

Photography For C4D Artists: 200 Free Tutorials.
https://www.youtube.com/user/DrSassiLA/playlists

NEW:

NEW: Cineversity [CV4]

Profile
 
 
   
 
 
‹‹ Soft bouncing models?      Symmetry issue ››