Thanks G. I must say i couldn't have done this without all the help from you and S. I'll keep updating my channel on youtube with the progress and see if i also can make links to them from the forum here.

I also have a shooting planned next week, but depending on what i can make out of the shots i got today, i may just cancel that, because it actually turned out way better than i had hoped for.

...

Great job! very well done set up using those ladders as tripods :)

That camera calibration looked great, actually better than I would have expected in an outdoor setting, one thing though, when using a very padded surface is, you will get compression on the pad, but the ground plane won't move, so it will throw off the feet to video perception.

A way to correct this is to use the actual hard ground for your light touches, then put the pad in for the recording background evaluation... the feet don't need to appear on the grid to track properly, especially for on pad compression, or you can try to disable the floor contact check box and the feet will penetrate the grid to follow the video, this is why you got some quirky feet actions also, but I would think it would work better using the hard ground for calibration, then if you need to you can raise the floor a bit before tracking, not much, just slightly to where the pad is fully compressed on the jump down, but adjust the actors height also accordingly slightly shorter.

One thing to note, on the tracking process it is always better with a human motion such as this to use the flexible spine option, NOT very flexible spine, that will most times cause a wiggly butt, even after JR, but you can try both ways with different JR settings and see, I saw you chose stiff spine, just re-track and use flex spine and select it before initial refit, this will bring the head and spine on refit and throughout the tracking into play more.

The foot color was a bit to close to the base pad color and the wall actually, more contrast is best, even the black shoes you shown before would work, the program actually tracks well with dark socks/shoes even if pants are dark color too, but you got about 90% tracking of them anyway.

For the first attempt that is working very good! ~~~Applause~~~

If you had the chance to work with Moves on top of the hands at some point, that would look incredibly realistic just in the iPi playback, although you can layer adjust the hand positioning pretty easily later, as there isn't very much motion on them.

Good work!

...

...

One other thing I noticed from your raw footage is, you may want to adjust your leg length to where the hip bone connects to the upper leg, it may work better to be positioned more to where the crease would occur in the mesh, not where you have it higher near your hip bone on the video.

This will help during the tracking to keep the leg bends more accurate and on squatting motions to keep the hips and back more in-line with the video.

Just a hint that may help improve the tracking quality, but if you feel it is fine as is, then that's your prerogative :)

...

Hi Snapz, always nice to get some cool feedback from you :-)

About the positioning, i was actually just experimenting with different settings for leg length, arm length and such to get all parts to line up the best. I think i'll post a video of how i position it, maybe i can still improve on it.

Already tried to run a tracking with flexible lower spine as per your suggestion... i'll let you know how it turned out when it's done.

Great ideas with the crash pads. I was actually thinking about how that would turn out, because it would obviously cause some change to the scene definition. When i do the next shoot, i need to drag out one of the bigger crashpads, because jumping down 20-30 times on that relatively thin pad did leave me with a sore back and aching knees.. so this will actually add about 25 cm of height to the scene. So i was actually gonna bring out a little foot stool (or whatever it's called) to stand on during the t-pose. The landing part is not really important for the footage i want to capture, allthough it is an interesting problem to see if i can tackle, so if i get time and got the energy for it, i'll get it dragged out, for a calibration at real ground plane as you suggest and see how that performs.

And about the footwear i agree, i brought the light boots but think i'll try with the darker ones. It does seem to blend it a bit too much with the grey background. On the other hand, the colored climbing hold did not confuse the tracking at all, i was actually kinda expecting it to mix it up with my hand, so next time i may just go crazy with some other color holds to see how that turn out.

...

Yeah, I figured this was your first run and you would try to adjust things on a next recording session where you felt were needed.

Those are just tips I saw from what you posted, once you recognize what you see and get out of many, many tracking runs, you will find out just where things need to be to get the best settings/results. Once you get your Actor parameters you like, don't forget to Save the Actor file for that session for later, easier use, but it may not work on a different recording session.

The little program is really robust with the PS Eyes, but over the time I have spent working with, I have found it very vital to have your Actor scaled properly, even if for some reason it isn't the exact height of you, or another performer that's used, but if the calibration is close and cam 1 height off the ground is set close to actual lens height, your Actor height should fall very close to actual and appear in scene just seeing the very bottom of the foot under the Actors foot after refit, with the heels just so slightly off the floor grid in the T-pose.

Don't worry that every cam is exactly the actual height off ground after cam 1 height is set in calibration, all cams will usually be very close to what the program needs to see to track just fine, but of course if it's way off after refit, then somethings gone awry after calibrating and if not caught, you will loose the whole recording session if you tear down the set up first, then find out, so you should at least test one video refit and partial tracking before tear down.

I always do one calibration before recording and one after session just in case, and use the best one of the 2, but my cams are permanently fixed, so I can run a calibration at any time after recording is finished.

Sometimes the tracker will just loose tracking for no real reason, if you are watching it track and see it, you can stop and go back to the last good frame, hit refit, then track forward again and the same loss won't happen most of the time, but sometimes it will, it's just one of the quirks in the program that happen, but always better to track forward when possible, than to track backwards to fix a spot when you can, and you can switch cam view temporarily to better track areas that won't track straight through on the primary tracking cams view.

Most of my dance, or action sequences are 30 sec to 1 min continuous, with a lot of motions, so I don't like to have a lot of clean up outside the program when I can do a lot of it automatically, and even make slight changes in settings at different points in the ROI first, it's just my way, not saying it's the right way, or the best way, just how I like to do things and I get good results, before and after exporting, so that's all that really matters to me.

You got very good results for just starting, so you can't go anywhere but better from there :) Good Work!

...

So im working my way through the mocap takes, which yielded a total of 20 takes. In this take 18, i experimented a bit with crossing hands and cross movements in general + matching holds. In general it turned out pretty well i think. Only thing i wish i could figure out how to fix is those jerky wrist movements. They keep turning for no reason from frame to frame, no matter how much i adjust and refit and track, so i guess this is something i need to fix in MotionBuilder anyway or is there a way to make this better?

https://youtu.be/V6MUDrYmaAk

I did one calibration before we shot and one after, both was close to perfect, but for all trackings i've done so far, i used the first calibration video. I may try and track a few of them with the second just to see if it changes anything noticable.

/Jonas

Here's the video showing the setup to get things into MotionBuilder and on to UE4 if this should be relevant for someone

https://youtu.be/zb8C-qXSjkU

That's looking really good considering the complexity of what you're trying to record.

Great job, and please keep posting! I'm looking forward to seeing how far you can take this.

G.

_________________
Greenlaw
Artist/Partner - Little Green Dog | My Demo Reels (2013,) (2015,) (2017,) and (2019)

Watch a one minute excerpt on Vimeo now!

...

As for the hand twitches, you will have that without using Moves data to absolutely control the hand, so manual correcting is all you can do.

iPi also has a pinning type feature if you weren't aware of this, if you select the bone at the good frame, then advance forward with the right keyboard arrow key, while with rotate tool selected and holding the mouse button down, by just slightly moving the mouse on each frame will make it snap back to the same position in each forward frame, (don't need to move it much), but don't release the mouse button until finished, or if you do release it, just go back a frame and repeat from that frame again, it is a good pinning type feature iPi has that many don't know about, or just forget it's there.

Much better results :) try adding some hand gripping files to the fingers, you can do this also right in iPi, you can make and Save your own or there are somewhere in the forum a hand gesture pack someone made up that has several pre-made files.

Maybe use search function in the forum to find it... type in Hand Gestures maybe?

You can also use MoBu to add hand gestures and save each file to a pose library for later use.

I did notice the UE4 Mannequins hands in iPi come in set a bit too far inward and even when Moves are used, I had to set the iPi Actors hands a bit further out to get desired results, so without Moves data, you might want to adjust the hands in a global layer change first in MoBu, or if manual posing them later will handle that.

Take looks good, very natural for the most part.

Also I think 1.82m should be the proper Actor height, since your floor height was a bit high and seeing your feet not matching as closely as they could once you were on the wall, which gives a pointed toe effect. Did you try disabling the ground collision on any tests before refit and tracking?

...

I did not know there was such a pinning function in ipi (excited) i need to go take a look at that for sure... I have those twitches on both hands and actually more on the shoulders too. I just recall looking at the curves in MoBu, and there were as much as 100 frames in a row where the x-axis jumped like 20 degrees up and then eventually found its way back to normal levels. Since manually adjusting curves is something i would do as little as possible, i will definetely see if that ipi frame mouse hold trick can iron out the worst of the misalignments.

Yesterday i did a very rough set of hand animations, by having the video file open in ipirecorder and watching the different angles on certain frames to see when i curled up the hands while matching up the frame in MoBu and posing the hands there.. It wasnt that tedius with the FK rig actually. I might bring in the different camera takes from each camera in MoBu to see if that makes it easier to work with.

I will however give it a few tries too, just to do the posing directly in ipi as you suggest.. The gimbals are just so hard to work with compared to MoBu where i can curl up all fingers by a few clicks and rotate.

Since the hands need to remain stationary at certain times of the clip, i guess i need to position constrain them as well in MoBo. Ipi doesnt have such a feature right?