I just spent the whole day testing yesterday and am very pleased with the results. :) I had some trouble with crossing my arms and squatting but I kinda expected that. My question is, Can I wear different colored gloves to help it discern one hand from the other? And like-wise could I wear different colored shoes too? Or does it sample one color for tracking? I could even go so far as custom tailoring (yay in-laws!) a shirt and pants with different colored sleeves and pant legs if this could help the program in difficult poses. From the sampling it shows now it looks like different color shoes and gloves could be an option and probably not the pants or shirt. Am I wrong and could this help tracking? Thanks. Now I need to go shovel the driveway for the Nth time today during our Blizzard in good 'ol Iowa. Cheers!

GR1F1TH

_________________

In general, our main strategy is supporting casual clothes.

Current version does not support different colors for left/right hands.

In theory, this can be implemented, though I am not sure how many users would want to use it. Currently, we concentrate on CUDA performance optimizations for new version, after that we may get back to adding more clothing models.

Thank you for the response. That saves me some time in testing.

_________________

I think differentiation between left and right color of suit is a very big improvement for iPi DMC, because the algorithm is based on the color of suit. Although it may be very difficult or even impossible to have such clothing in the market. But it would be much easier to design a special outfit like that instead of having to constantly correct the tracking error between the right hand and left hand and also between the right leg and left leg. It's an exhausting job and spent a considerable amount of time.

Iwan Suryolaksono

Making a suit with different colors shouldn't be hard. Take white clothing and add whatever color clothing dye you want to whatever part you want. You could easily make a suit with a different color for each body part if you wanted.

I agreed with the right/left improuvement, i'm sure it can help during slow motion with no large amplitude, like a discution for example

_________________
www.apside3d.fr

Such has been done at MIT for hand gestures...


MIT’s Ridiculously Colorful Glove is Latest Hand Tracking Interface

MIT unveils colorful solution for cheap, accessible gesture-based computing

yeah I wasn't talking about actually tracking fingers and such I just wanted better tracking of which wrist was which when they cross over or by each other. that is very neat stuff but it based (or was back when I read about this) off a library of pre-cached hand positions.

_________________

That is an interesting point. It does attempt to do a database image match but it doesn't say if that is for relative position or absolute position. I assume it would be for relative position or else interpolation would come in effect.



I hear ya' I only brought this up as one solution that has been boasted about being a cheap way of increasing accuracy. I believe most optical motion capture solutions use some sort of marker that enables the software to track more accurately/efficiently. OptiTrack uses reflective markers and PhaseSpace uses active LED markers for their optical capture systems. Both use black suits to create the greatest amount of contrast for marker recognition.

I use Zign Track for facial motion capture. There we use color (you pick a single color of choice) dots applied to the face for tracking markers.

Your are using Zign Track?, Is it possible to use it with 3dsmax, and How ?Wich version is tit necessary to use and is difficult to use it ?

_________________
www.apside3d.fr