I’ve been using optical flow in batch mode for some time now and it’s amazing. One small suggestion or rather a question would be, if it’d be possible to use the last frame as a starting point for the next one. For example, if I had a series of scans of an actor, gradually opening his mouth, with the first frame being completely closed, and the last one fully opened, I’d want wrap to always take the previously processed frame and start with that instead of the original input.
Not sure if it’s possible to put it in as an extra option or if it’s too much of a hassle, but it’d be really handy. Also, that way I could make point selects just for key frames not the entire sequence, since that’d take quite long, I’m processing hundreds of scans at a time.
We’ve been experimenting with 4D for a while and have managed to process some sequences using the following approach:
We look through the sequence and find keyframes in it;
For each keyframe we process them very carefully using OpticalFlowWrapping and manual cleanup;
We put the keyframe meshes on a timeline in Houdini and animate intermediate frames using simple mesh blending. It provides a rough approximation of the really sequence.
Then we fit those approximated meshes to the scans in each frame using OpticalFlowWrapping. Since the meshes a fairly close, it doesn’t require control points. One advantage of such approach is that it does not suffer from error accumulation. If there is a noise, it’s really random and can be easily filtered.
thank you for the pointers! That’s very similar to what I’m doing at the moment, only using a different package for assisted tracking (and very long shors). I was afraid error accumulation might occur, so this still might be the best solution. I was just trying to think of ways to automate as much of the process as possible.
That’s exactly our goal. We are thinking of adding new tools that will automate the processing of FACs scans and 4D sequences as much as we can, but I don’t know how much time it can take.