Yes, Lopolo, precision is everything in this case.
I remember when Dr. Steve Baines showed his ideas of how to create a motion tracker during a meeting in London back in 2005. He used little sticks to introduce this to the group (PXC digital artist group). When you take two sticks and have them crossed where the tracker point is, and the angle between the two is equal to the camera move, you get often a very small angle, which results in a very imprecise crossing point location. If those sticks where 90º to each other, the crossing point would be super precise.
The longer the lens, and the smaller the sensor, while the movement (object or Camera, perhaps both) is also very little, the sticks would be nearly parallel, and close to be un-solve-able. Hence my suggestions.
To use a 360º shutter angle isn’t helping here at all of course, as you correctly mentioned. But if you need a video look, then you might do that in post, (RE:Vision Smart Blur), as the parts than needs to be placed into the shirt, need motion blur as well. It is easier to get believable results if that is done in one step, than to match an existent blur, at least that is my take on it.
Yes, Lopolo, precision is everything in this case.
I remember when Dr. Steve Baines showed his ideas of how to create a motion tracker during a meeting in London back in 2005. He used little sticks to introduce this to the group (PXC digital artist group). When you take two sticks and have them crossed where the tracker point is, and the angle between the two is equal to the camera move, you get often a very small angle, which results in a very imprecise crossing point location. If those sticks where 90º to each other, the crossing point would be super precise.
The longer the lens, and the smaller the sensor, while the movement (object or Camera, perhaps both) is also very little, the sticks would be nearly parallel, and close to be un-solve-able. Hence my suggestions.
To be honest, I didn’t get it. Are we talking about 2D tracking? I think the crossed markers are the best for this type (except 3D pyramid) of shot because you can easily find and guess the centre of the marker even the shot is blurred. I’m talking about handmade corrections. Don’t you agree? Do I miss something?
Dr. Sassi - 25 February 2019 02:52 AM
To use a 360º shutter angle isn’t helping here at all of course, as you correctly mentioned. But if you need a video look, then you might do that in post, (RE:Vision Smart Blur), as the parts than needs to be placed into the shirt, need motion blur as well. It is easier to get believable results if that is done in one step, than to match an existent blur, at least that is my take on it.
I don’t believe in post production in this case. It is similar to post DOF it doesn’t work much. But yes, sometimes this is better solution than none
The 2D track is the first step. Object tracking is a 3D process. Cinema 4D is not able to create a four corner pin tracking “effect” for example, which would be possible already with 2D.
All the 2D tracking features are used in 3D space, they are evaluated as a position in space per frame, which results in a single axis animation for each Object Tracker.
I was not talking about the shape of the tracking marker here. I assumed that you like to have some 3D on your short or inside of the body. The crossings are based on the motion of the tracker, like viewing rays.
However, you can track anything and use whatever result you get to do anything you like. Perhaps even create a four corner pin effect in post.
Yes, you have to decide what is your workflow. I can only share what I have learned and used over the past decades doing this. If you like to avoid the post, that is, of course, your decision.
The 2D track is the first step. Object tracking is a 3D process. Cinema 4D is not able to create a four corner pin tracking “effect” for example, which would be possible already with 2D.
All the 2D tracking features are used in 3D space, they are evaluated as a position in space per frame, which results in a single axis animation for each Object Tracker.
I was not talking about the shape of the tracking marker here. I assumed that you like to have some 3D on your short or inside of the body. The crossings are based on the motion of the tracker, like viewing rays.
However, you can track anything and use whatever result you get to do anything you like. Perhaps even create a four corner pin effect in post.
Yes, you have to decide what is your workflow. I can only share what I have learned and used over the past decades doing this. If you like to avoid the post, that is, of course, your decision.
Cheers
Ohh .. my poor english. Im sorry, it was misunderstanding. I translated crossed sticks as a markers which is my bad. This one “When you take two sticks and have them crossed where the tracker point is” confused me. Shame on me
Great, Lopolo, try the suggestions from above. With all the little x tracked carefully it should be better than my quick attempt to get it to work. Yes, to track all of them takes time.
The image below shows what I had in mind with the pyramid. It can be just folded paper glued on the cardboard. With some extra markers. Use more different markers than the only x, to avoid the frequent jumping of the features, as they all look at some time the same. Perhaps use even some color, as that color is not destroyed as quickly based on your 360º motion blur.
Image https://www.amazon.com/clouddrive/share/iYAqGkaXBPitWmStA0EiNRYq9s7EA2LdgWusURn8v4o
I tried to use the cardboard as the only source and pin so the camera to the board (as if the board would be static and the camera moves.)
Similar to this here: https://www.cineversity.com/forums/viewthread/3149/#12100
But the area is just too small to get anything better than before.
As mentioned before, the area is just tiny, and without a lens-profile and tracking all possible points (to minimize the error) I have little hope that the result will work. Even if all of that is done, I’m not confident to get a solid track, but that will be known after you went through this.
Preparation is one of the keys in motion tracking. There more spatial and parallax (changing!) information the algorithm can retrieve from the given shot, the more the model is likely to be created. The model could be the object or the environment. After all, tracking tries to find information in space. All given tracking features have some kind of error in it, especially given that the footage received is often created (Bayer Pattern)and not really based on Photosites equal pixels, like with a three-chip camera back in the days. After all, points are spatially defined, the evaluation of how each point relates to the “cloud” of points will allow filtering the more “independent” points out. After all, the environment or a solid object has no movement in itself. The group of points with the strongest agreement among each other, create the static cloud of positions. This is then used, to calculate the camera or the object. If the objects bend, that calculation is flawed.
Keep in mind that fast movements can create its own problem, as CMOS cameras (vs. CCD) often have more or less to deal with Rolling-Shutter problems. Which means a fast movement to one side can ‘Shear’ the result.
Anyway, I think I have nothing to add to this shot. I hope the tracking will result in something you can work with, despite all the limitations. But that is what makes testing so great, you can explore the boundaries.
This test was intentionally hard task to get better idea where are limits of build-in tracker of C4D. Yes, Im aware of preproduction part is crucial and I know limitations of CMOS known as rolling shutter etc.
Yes, the tracking plane is not rock solid but it was part of my test. The first idea was to make more individual planes instead of one and then join them together to get final bending shape. As you said, the cloud of points have to go through the algorithm to eliminate od adjust more independent points which had impact to all of them I guess. I wanted to try to get few points only but double (or more) precise rather than “average” of point cloud. Of course, i have to use lens correction first.
This is not commercial take, just a test and I agree, testing and exploring is fun
Btw, I wish to have oportunity define static tracking point (as simple 2D position - but maybe it is there already, i didnt saw all corners of tracker yet), not to lose data when changing sequence and few other features in future version.
Thank you once more for your time and advices! It was helpful!
Your test was well tuned, pretty much exploring around the edge. Those tests show how things work together, and how quickly it can lead to unsatisfactory results when one element in the chain is not at its best.
I’m confident that many others might find something here that points to parts of their workflow which might have some space for improvement. For my part, I always assume that I miss something or that some weird workflow might make it happen anyway. I have indeed invested some time here, with the hope that it will have some benefits, as well as beyond your exploration.
Learning Syntheyes via FXPHD, (after learning other stuff in the PXC before that, etc.) from 2007 on was certainly an eye-opener, and the number of workarounds and tricks I have seen was breathtaking.
Regarding your suggestions, the best place to do so is to drop a message here: https://www.maxon.net/en-us/support/suggestions/
It is undoubtedly the fastest way to get an idea onto the wishlist.
I say thank you for your patience and endurance along the way so far. I have the feeling that it was not the last question. Please never hesitate to ask, the earlier (in terms of pre-production) the better.
My best wishes
P.S.: I saw your reply, but we had the web down, and my phone’s two browsers don’t allow for answering here, I hope an update will fix that one day.
I really love to double check everything, at least what I’m talking about here. Quoting from memory isn’t really something that I found to be sufficient in these days, where everything is changing so fast. However, tracking can be challenging or straightforward. It is not really the tracking application that is in the way, from my point of view, it is the information we prepare and share. So, pre-production decisions are essential, and that means knowledge. I hope I have shared enough above to fill some gaps that might be or not.
I picked a little part from the clip (frame 70-120) and used every single little X on the card, as well as tracked all four corners. The 70-120 was used, as it has one of the more heavy motion-blur in the middle.
So, another few hours into it, based on 64 (!) manually tracked features, frame by frame, here is the result.
Amazing. I knew it is possible and you brought a proof You chose the most difficult part and even SS was 1/25, higher focal length, low angle etc., the result seems to be very good!
I think we agree each other and yes, the pre-production is essential. However, sometimes there is no posibility to get higher shutter speed when director wants to focus on movement or to change the composition of shot to get easier postproduction job afterwards. These two requirements are often contradicting each other so better be prepared for non ideal conditions. VFX artists point of view is not alone in entire process and for instance, they want to have far more light on the scene than director or cameraman.
btw, Did you try to use 8 corner points only? I wish I would have some free time this week to try it as well
I prefer with my camera a 180º shutter, which is just the most cinematic to my eye. A video is more 360º, not my thing at all. I even bought the Motion Mount element for my Red Epic Dragon, to have more control over it. But I understand, if the director feels like knowing Cinematography well, things are done differently. (I use electronic/digital cameras since 1986, so I might have some ideas about my own style)
In VFX, it depends on contracts, and who has the final say, I agree. If a shot costs ten times the amount to ‘fix-it-in-post’ than to do it right from the beginning, things will all of the sudden work differently, … other than if the VFX house has offered a flat fee for a shot.
Anyway, I love to dig deep and check everything practically. Just quoting other people doesn’t really work for me. I guess that is similar on your side, hence your test here.
