Motion Tracking & Object Tracking inside Cinema 4D: The Good, the Bad, and the Ugly Trackers

Share this video
  • Duration: 04:53
  • Views: 3899
  • Made with Release: 18
  • Works with Release: 18 and greater

The Good, the Bad, and the Ugly Trackers

The ideal shape for our 2D tracking algorithm would be a square noise pattern that doesn't change size, orientation, or perspective. Unfortunately, because you won't find these ideal trackers anywhere, this video will show you what good and bad trackers are.

Less...

Transcript

In this video I will show you what is a good tracker, what is a bad tracker, but I will show you nothing about ugly trackers. I just wanted a catchy title. The ideal shape for our 2D tracking algorithm would be a square noise pattern that doesn't change size, orientation or perspective. Unfortunately, you won't find these ideal trackers very often, or at all, for that matter. In real world scenarios, you only really need to concern yourself about the manual trackers, as the automatic trackers will find their own patterns to sit on. As far as the manual trackers are concerned, we set the manual tracker and then press a button to tell it to track. When the track loses its focus, we correct it, rinse and repeat. But here's the not so good news. There are times when you will have to adjust the position of the manual tracker on every single frame, for quite a few consecutive frames. So having a shape that allows for accurate placement without too much speculation, is necessary. In my book, small spheres and pointy things are perfect for that job. The spheres, because regardless of their orientation, always look like circles in two dimensions, allow for the placement of the pattern square aligned to their bounding edges, and that tracker will give us the center of the 3D sphere. This is ideal when using them for object tracking, as you can choose those centers to align your 3D objects. The pointy things, on the other hand, are ideal for tracking larger areas, green screen, backgrounds, etc., as you can center the tracker on the tip of the point with relative ease. Be prepared, though. Sometimes, seemingly good manual trackers will fail to track miserably for reasons that seem unintelligible at the time, but are actually very logical. A change of size or orientation, perspective skew, change of contrast and other things, do make the tracked feature look very different from the original one, and can therefore throw off your tracking. In those cases, a deep breath and some fully manual tracking will get you going with patience and tenacity being your best tools. The good thing is that seven good trackers will get you a very good 3D track. That's the absolute minimum. Add a few more for safety, and the overall work shouldn't be that much. But be prepared, nevertheless. The most important thing is track quality. Ten very good tracks will yield a great 3D track, whereas 100 not so good tracks will fail totally. Talking about not so good tracks brings me to the next subject. What are bad features? The most common cases are the following. Reflections and speculars. Often, bright dots and patterns caused by reflections and speculars can fool the tracker, as these tend to move with your camera, rather than fixed on your surface. Obviously, these tracks won't help the solve. On the contrary, they will confuse the tracker and sometimes lead to bad solves. Beware, these little buggers show their face more often than you think. Wooden tables, for example, have a varnished finish, but are bumpy. Any reflections or speculars will produce very jittery tracks. Second, we have overlapping objects. Sometimes, what seems to the tracker to be a nice pattern to follow is not a fixed point in space, but the result of two different objects overlapping in 2D space. In this case, the tracked point is totally useless, as it doesn't represent any usable information. It's just a visual artifact. Lastly, randomly identical areas. Sometimes the tracker can be fooled, and jump to a close by area of the image, because there was some similarity in the pattern. Just like above, these can throw off the solve. These jumps can happen on consecutive frames, so you may see the tracker bouncing here and there. That's a very bad track. There are remedies for all of the above. Bad tracks can be identified most of the times, using an array of provided tools we will see in detail later on. Now you should have a rough idea about tracking and trackers. Knowing these things will definitely make your life easier when tracking. And since I was talking about making your life easier, in the following video we will see what a survey shot is, and why we need it.
Resume Auto-Scroll?