 |
|
||||||
|
FAQ How do I make a template for solving multiple takes? How do I import motion capture data in to the template? Is it possible to use peelSolve as part of a renderFarm/solvingFarm? I have run the solve but the spine looks terrible, how do I fix it? Is there a specific markerset I should use? Why does a bone keep popping or moving in a unnatrual way? Template for solving multiple takes The template should be built from a tpose and should have the motion capture hierarchy and the skeleton hierarchy with the markers attached. Delete the animation on the motion capture hierarchy. Add joint attributes to all joints. Set the root node and any solve options. Save this scene off and use it as the starting point for doing solves. If you find a common problem that keeps coming up while solving a number of moves, identify the problem and apply the solution/workaround to the template file. Avoid importing data, solving, saving then importing new data in to that scene, load the template up again. For working on multiple takes, select the top node of the c3d hierarchy and turn on "merge" in the c3d or trc importer. Alternatively import it as a new hierarchy, select the top node and use "Reconnect mocap root" in the peelSolve dropdown menu, to connect the markers to that hierarchy. Both techniques do a match on name - the second one matches on the markers that are parented to the skeleton (removing "_marker"), the first matches on the motion capture data's name. The fastest way to set up a solve farm is open an instance of maya, load the plugin and execute the solver using the mel command "peelSolve". There is provision for exporting out an xml file that can be sent to a standalone command line application, but this has not been fully implemented or tested yet. Getting good results for the spine The spine can be tricky to solve. Having extra markers on the back really helps to get a great looking result for the spine. If you are using the "lite" version of the solver you will probably have to reduce the number of spine joints or lock the rotation channels on joints that are not being driven by markers. If you have the "pro" version, setting rotation stiffness is essential for joints that are not being driven by markers. Also turning on lendof and setting lenStiff to 1 for the spine joints to allow some translation in the spine can really help loosen up the system. PeelSolve does not require a specific markerset, but having extra markers in certain places such as near the spine can help the solver get a great looking result. 3 markers for the feet, hands and head is essential for getting the orientations of those elements. One extra marker for the toe is optional. 4 markers for the root is common, 3 minimum 1 marker for each of the elbows and knees is fine, having more on the arms or legs can help to identify problems 4 markers for the chest/ribcage is useful, 3 minimum. Should be aligned with the spine. The markers driving the spine and ribcage should be symmetrical. 4 markers on the lower back forming a box about 3" apart from each other is useful for driving displacement of the lower torso, e.g. for "folding" the spine. I am trying to get people to call these the "Macleod" makers :) 1 marker on the shoulder is okay, 2 is better. Avoid putting markers on the shoulder-blades, keep the markers as close to the round part of the shoulder as possible. Two markers give better results for the clavicle. Asymmetrical markers are often added to the torso to help with auto-labelling and realtime identification of the markers. I'd strongly advice not to use those markers in the solve. Keep the markers symmetrical on the torso otherwise you will likely get some asymmetrical looking results as the markers push or pull in different directions. Unnatural movements are caused by the skeleton system being driven not representing the original capture subject (joint center positions) or the markers not being placed correctly in relation to the joint centers. Finding the a solution to the problem requires an understanding of the movement you are trying to apply and the system you are trying to apply it to. Basically, something isn't matching up somewhere and it's creating an artifact. You will either need to match things up better, or hide the artifact somewhere. The best place to start looking for the problem is to see where markers are either falling short or going beyond their targets. For instance if the feet are lifting off the ground when they shouldn't be, most likely you will find the the markers are falling short of their targets both at the feet and at the hips. This is an indication that the legs are not long enough and the root is pulling the feet up. The solution would be to make the legs longer, e.g. to bend the knees slightly for the initialization pose. If the clavicle pops while the arm is swinging and the arm markers are going past their
targets, this means the arm is falling short of the target and is being pushed in to the
torso, causing an artifact. The solution could be to turn the elbow marker in to an
"aim" marker rather than a pos, to reduce the pressure on of the arm pushing towards the
torso. This will result in an incorrect position for the elbow, but will "hide" the
artifact. Alternativly the joint center for the shoulder and elbow can be moved to a
position that allow for the motion of the models arm to move more naturally. Even small
changes to the position of the joint centers can have quite dramatic effects on how
well the data can be applied to the skeleton.
|
