Join Andy Needham for an in-depth discussion in this video Solving the camera data and refining the results, part of Creating a Show Open in Nuke X and CINEMA 4D.
So, we've tracked our shots. We've removed points in the sky that aren't going to help the solve. Now we want to see what kind of solve value we get when we do our initial solve. So, we're going to press solve here and that will solve the cameramation. So settling the camera just means we're taking the tracks that have been tracked in the tracking phase and producing a 3D representation of the camera used to shoot the original footage. So our initial solve error is 1.49.
Not too bad, not too bad to start off with. You never know, you'll definitely see higher areas than this, and that's usually a sign that you've either got your track settings wrong or, you know, you've either masked out too much of the shot. Don't be afraid of going back and retracking the shot if need be. You know, you might have to throw everything out and start again. But for now we are going to reduce this error. We want to get it below one for HD footage. You know, if you've got 3k, 4k you can probably get away with a higher error.
And its kind of relative to the frame size. So certainly for something like full HD. There you will want to be under one. Otherwise, you're going to see slipping in your tracks and you know, nobody wants that. Before we solved the short, we saw that there was some points that had really weird motion. I can just press space bar to max the view here. And we talked about how camera tracker would just reject those points. It's done so with this one. It just doesn't like it at all. And you know, we can see why.
That is just not a good point. What we want to do now is use threshold and the. Various settings in the autotracks tab to go through and pick out these points that have large errors and just remove them from our solve. So, the maximum error that we're coming across at the moment is 8.86. We can definitely get that lower. If we click into the error max field,.
And hold down command, and then click the max error. If it's not already viewable here, press F just to frame up the graph. So the max error is the threshold for the error max. And the error max if we hold over here is the maximum reproduction error per frame. So, this is the representation of our shot. We can scrub through it. It's like a play head. And you know that if we move to this part of the shot, this is where the higher areas are occurring.
So, we can use the thresholds to actually bring this down. And you can see, if I move the threshold slightly, it's. Starting to reject points. If we just frame the viewer, we'll move along in time, and somewhere around here, and if I bring down the threshold drastically, you know, now we know that a lot of points are rejected. So, we want to start bringing this back. To an area where more of the good tracks are coming back.
And one thing to bear in mind and this is why we tracked so many features is your number of tracks. As we go through culling large numbers of features, we're going to reduce the number of tracks throughout the shop. And you want to be careful not to go below one hundred feet just because the camera tracker needs that to create an accurate solution. At the moment we're pretty safe, you know, we've got thousands of features on the frame so far. So we definitely have enough room to go through the shot and cut large numbers of these features that are just Producing huge errors.
So we'll do that, initially. And the workflow is, as we've done before, select your error max. And your max error. Frame up the result. Reduce the threshold. Delete the rejected points. Will ask you if you are sure you want to do that. And we'll say yes. And then it's going to think about it for a second, and it will update the solve, and you can see we've actually reduced the error. So, you should go through your shot, reducing the max error, but you can also reduce the max track error as well.
If we frame that up and this one will be quite sensitive, so you might want to just use the numbers by clicking into the field and pressing the down and up arrows, you can do that, and if you go to the right of a number, you go into a different increment. So, again, same work flow, reduce the threshold and delete the rejected points. At the same time I'm just going to actually increase the minimum length of the tracks to say lets actually use the threshold here in length and track length min.
And if we see we've got we can move this quite away. So I'm going to make sure it's around the 20 mark along the tracks are going to be more reliable so we're going to delete the rejected points and say yes to the warning. And even though it's a small change it's a change in the right direction. So what we need to do really is continue working in this way to bring the error down. So, as we've seen, this process is a lot of tweaking, and, it will take some time, just to get the arrow down.
You will need to, as said, bear in mind the number of tracks, that you're not deleting too many. So, it's a slow and steady process, and I want to spend some more time to get this arrow down below one before moving on to the next video. So, if you're following along, you should go ahead and do that too.
- Setting up the CAMERATRACKER node
- Solving the camera data
- Setting a ground plane
- Testing the solve with temporary geometry
- Randomizing particle motion
- Lighting the scene
- Compositing in Nuke