Setting up ps3 mover
Page 6 of 12

Author:  funwithps3 [ Thu Nov 01, 2018 12:46 am ]
Post subject:  Re: Setting up ps3 mover

Greenlaw wrote:
Did Kinect v2 sensors come down in price? For a while, the prices got really crazy when MS announced discontinuation. Luckily, I had purchased 4 of them back when they first came out so I'm set, but I wonder if should get another backup if they're 'cheap' right now. I'll check this evening but what's the typical price right now?

TBH, I'd rather look into something new to replace my v2 sensors if and when they eventually go bad. Unfortunately, most of current crop of alternative depth sensors are apparently comparable to v1 sensors, not v2.

Has anybody tried any of the higher-end Kinect-alternative sensors yet?

Thankyou Greenlaw !
Well bought mine used kinnect v2 50 USD 3 month ago and thats a normal price on the site tradera..but they dont ship abroad :-(

If you mean the new Kinect version from microsoft i have not seen them on the market yet, ?

Best regards

Author:  funwithps3 [ Thu Nov 01, 2018 1:10 am ]
Post subject:  Re: Setting up ps3 mover

Snapz wrote:

As to the alternatives G, they are kind of expensive, some in the range of a new Kinect v2, some even more pricey brand new, used if can find them would be cheaper of course.

You can look at the comparison videos of each on YouTube and correct, none match the point cloud of the Kinect v2, which is also best recommended by iPi.

To the jittery hand issue of previous poster, the PS Eye full processing path must include the running of the "Refine" process, either forward, or backward for the entire ROI you are tracking and the Move data must be applied always as the final step, before playback review, or export of the animation.

This process should also be run even with Kinect v2 tracking for best results, but many leave it out, as it takes extra time to run it, though it is much faster with Kinects.

PS Eyes are more of a Prosumer set up, things should be much more tightly respected, as far as the room, lighting, clothing, space, colors of all corresponding parameters, just much more consistently followed, not just a plug-n-play set up as easy as Kinects, and even if all parameters are followed the calibration must be very good and the iPi Actor set up correctly, but once this is all realized it is an excellent solution for mocap, and I would put my animations up against any other budget solution available now, though it isn't a real-time solution, but in my personal experience far less clean up in the exported animations than most real-time solutions, especially now with v4, for standard sized human figures anyway, it gets a bit tricky to act as if you were a non-standard character, but not impossible, if you know in advance what character you are performing for.

Of course not all users want to put in the time to get good results, they would rather get an instant recorded export and spend hours upon hours cleaning it up, more power to them, but I handle 90% of this right inside iPi Studio very quickly before export, especially since the recent updates to fix the upper arm twist issue, but of course you have to know what to look for and how to fix it to do this efficiently, and I at least know my motions are very close to what was performed, because I am tracking and can see the actual video reference while running it through.

Until you really satisfy all the necessary parameters to run a PS Eye set up efficiently, you are just going to be disappointed in the results, that is why I believe you would be better off with dual Kinect v2, even though it takes a compatible 2nd laptop to run a second sensor, but even that set up has its limitations.

If your calibration is well done, you should really only need to align the iPi Actor in one camera, usually camera 1, or the best viewing camera anyway, then on refit it will snap into proper position in all cameras, if it doesn't, then the calibration is off, even if you got the notorious "Perfect Green" showing.

It is very hard for someone else to explain, so it is really just a trial by error until you figure it out to get the best tracking results you can with a PS Eyes set up and some users just have better luck than others.

With the new tracking algorithm, it is much more accurate and full tracking loss should be at a minimum with PS Eyes, so if it isn't, there is something not quite right that you will need to track down and fix, that's just how it is, or you can use another solution for mocap and deal with a whole different set of specifications for its performance.

Just starting out with this program, or any other, you need to learn the curves for better results, many have been explained throughout this forum and can be searched by issue with the search function in the top ribbon.

Not much more can be said, especially for the area you are trying to use, but if you stick with it and follow the correct requirements it does work well, in my sole opinion of course.


Thankyou Snapz !

Today ive loaded my ip video file from yesterday and after som changes in the actors right hand position and other parts of his body i did a new track, jittering process and then movers attachment ( which i should have done before jittering process).. but now his right arm dont shake so much… but i can se during the tracking process that his right arm dont follow exactly in a few seconds.. and that result is not so good. Some more practice on placing actor and postprocess , the better is the sequense… so im rather satisfied. There is one thing.. when i place the actors hand i rotate them to fit but when i refit his pose under actors tab... they goes back to the org postion . Wonder why. After attaching the movers they are in the right psn.

So i also tried the refine process but now there is a lot of michmach . The hands doesnt follows at all... well ive seen this before and i think there is a trick but i dont know which ?

Ive also tried with the new algoritm and older one. With the new one my gpu gtx 980 ti runs up to 98 % and thats good but the result vs older algoritm is not so good so i use the older.

Well Im that kind of person that never gives up.. so i feel that i will manage to get good results in a few weeks :-)

Best regards

Author:  Snapz [ Thu Nov 01, 2018 6:22 am ]
Post subject:  Re: Setting up ps3 mover


Don't worry about the hand position while in track forward, it should refit close enough to the correct position for tracking, and yes, it will snap to straight with the forearm while tracking forward always, if you start in a T Pose, keep the palms down and the thumbs at a natural angle, slightly downward.

You can add the Move data at the start frame before you run the refine process, or you can do it after, I do it before just out of habit and it will retain the wrist rotations while refining.

You should be able to use the new algorithm for tracking without many issues, but again, this relies greatly on several other factors you may not be meeting fully, so some tracking loss may occur due to that.

95-100% GPU usage while processing is exactly what you want and with a 980 Ti you should be near 3 fps tracking and close to 3.5 fps during refining, using the new algorithm if using the low resolution speed, is this true? If using high res speed tracking, which you probably are at this point, you will loose about 1 fps, but you will loose much more fps speed when using the old algorithm, but if that's what it takes to get better results until you get the hang of the program a bit more, then it is what it is.

As I stated before, the use of PS Eye camera is much more strict in following the parameters needed and must be followed consistently, you should be aware that the iPi scene light is also needed to be placed properly, as iPi tracking uses directional lighting and this should also reflect in the lighting of your capture area of your room with enough light for all cameras to see the colors of the Actor clearly and a brighter light coming from a single direction onto the Actor, while still trying not to cast harsh self shadows, or floor shadows, this is why the use of more lower wattage light points around the actor is better than less high wattage bright lighting from fewer points.

I am not going to say it is easy for every user to accommodate the needs for PS Eyes for best results, but if you stick with it long enough and consistently record with the proper settings and set up, eventually you will get better results, but without following the needs for that system, it may frustrate you more than convince you that good results are possible.

Many don't like the tracking speeds with PS Eyes, as it is a longer time process, but personally I feel, since I get good tracking results, the time spent processing is less than spending hours cleaning up poor real-time mocap from other budget systems, since most of the clean up can be done easily and basically automatically right inside iPi Studio for the most part.

You should probably stick to shorter, easier motions at this point and not long drawn out performances until you get a better feel for the program using PS Eyes and Move controllers, then work your way up to faster and more complex actions and get yourself a consistent process worked out that works best for you, but there is a path of the processing that should be followed for best results and then stuck with.

First, work on getting the best calibration you can, this is critical to the best tracking results and once the video and cal files are opened in Studio, it is best to spin the floor grid square to the Actor first, before refitting the Actor, this will help when the IK Move/Rotate tool is used later.

Second, is to get your capture area and lighting much better and your camera positioning better, aligning them all to one single spot on the center of the floor, and best to use the view port grid to achieve this with the center point of the grid being at hip level, or as close as possible to that.

Third, is to wear the proper contrasting colored, tighter fitting clothing, even light material darker colored gloves and a darker head wrap will help greatly.

Forth, is to set up the iPi Actor scaling and skeleton properly, lowering the leg length to get the hips a bit lower to center of mass usually is required, and the sizing sliders for the overall body mass should be set low, then try to match the rest of the sliders accordingly to be slightly thinner than the actual performer for better results.

Fifth, run track forward with whatever processing works best for you, you may need to stop and re-start the tracking at some points, this is common, but shouldn't be excessive tracking losses in my opinion, or something else may be wrong.

Sixth, run the refine process, either forward, or backward, with or without applying the Moves data first, and if you see an area come up where the tracking was off a bit, you can stop and track forward back over those areas to help clean it up, BUT DO NOT HIT THE REFIT BUTTON WHEN DOING THIS, or it will erase any prior run frames that have had refine on them, just track forward, or backward, then re-start the Refine process a few frames prior, or after the spot you stopped to re-track, depending on which way you were running the Refining.

Once you get to this point, re-apply the Moves data at the ROI start frame for the take, (you must rotate the Moves on the Actor to the correct position first using the Move Tool), and if the hands aren't positioned closely to the video hands position, rotate them better and re-apply the Data.

Then you can replay the animation at this point, it will be a bit jittery yet, but you are just looking for any glitchy, or bad areas that may need more refining and take care of them in this phase.

Once all is to your satisfaction, you can run the Jitter Removal process, but best not to go above the default "2" settings per body part, except the head you may have to go higher to "5", especially if not using a head Move.

Finally, just let your animation playback in a loop several times and watch it for any areas that you might want to address, then scale the ROI only to encompass that area, so you don't mess up the entire ROI accidentally, because any changes, or re-tracking of those areas will need to have the entire process re-run, at least from the Refine portion forward, if not to include the re-tracking forward in the area, there is a few tricks to master when doing this, but you should be able to figure them out.

After all this done and you are happy, remembering the Moves data must always be the final step after any repairs and definitely before final exporting, whether you use finger gestures from inside iPi, or after in another 3D editing package.

In my experience with iPi, it is better to export your animations with the trajectory filter set high, 4-5, to pre-smooth the animations better, this is my personal preference, but is user choice, and not necessary if you want to smooth the animations in another way in post, depending on your 3D editing package, but it does help to pre-smooth it with iPi first.

I tried to lay out a basic processing path, so I hope it helps some, though no one can guarantee individual user results.

Good Luck and keep things posted for you progression.


Author:  Greenlaw [ Thu Nov 01, 2018 11:50 am ]
Post subject:  Re: Setting up ps3 mover

I knew I could count on Snapz to fill in the holes in my info! :)

Snapz wrote:
You can look at the comparison videos of each on YouTube and correct, none match the point cloud of the Kinect v2, which is also best recommended by iPi.

Thanks for the tip. I'll check it out this weekend.

PS Eyes are more of a Prosumer set up...

In general, I agree, except for the fact that I've used the Kinect v1/v2 setups to professionally create visual effects in a few 'SyFy' feature films, mostly for digital stunt double animation and creature effects. I'll concede that the quality is not as good as what I can get with PS3 Eye but, TBH, mocap data doesn't always need to be perfectly accurate for professional use. This was especially true when I was working freelance assignments with ridiculously tight deadlines. In these situations, getting convincing results quickly by recording in my living room with multiple Kinect sensors, outweighed spending more time getting more accurate performances by recording with PS3 Eye at an off-site studio location. If it looked good in context and got done ahead of schedule, neither I nor the clients cared if the data wasn't 100% accurate.

Anyway, for new users with the impression that Kinect is 'not professional', I personally would place Kinect in the 'Prosumer' category as well...just on the lower end of it when compared to PS3 Eye. This is because, IMO, there's a big quality distinction between iPi Mocap Studio's use of Kinect/depth sensor tech, and what you get with competing mocap software that also use depth sensors.

Author:  Snapz [ Thu Nov 01, 2018 12:53 pm ]
Post subject:  Re: Setting up ps3 mover


When I mentioned "Prosumer", I was more referring to the set up needed, (lighting, clothing, capture volume size, the set up being more strict), not actually the quality of the captures, or tracking, though there is a difference there also, IMO, but it is fairly cleaner and more usable than Kinect v1.

Dual Kinect v2 just didn't cut it for me, I never tried 3, but I doubt it is much better quality wise overall, but that is personal opinion, not a condemnation of the Kinect v2 set up, but now for Kinect v1, you might as well use that as a paper weight, as far as I am concerned :)


Author:  Greenlaw [ Thu Nov 01, 2018 1:25 pm ]
Post subject:  Re: Setting up ps3 mover

Fair enough. :)

Snapz wrote:
Dual Kinect v2 just didn't cut it for me, I never tried 3, but I doubt it is much better quality wise overall...

Sigh! By now, I was hoping to have at least 3 Kinect v2's up and running (my original intent was to have four running.) A few months ago, I finally added a new computer that can easily handle v2 capture but that was to replace the workstation that died on me earlier this year. D'oh! So, still capturing with two v2's for now.

I'm thinking of repairing my dead workstation so I can try 3 again but I'm not sure the cost will be worth it. It was pretty old though so it make more sense to chuck it and get a refurbished newer system. TBD.

As for quality with three, I think adding a third v2 will just help prevent more occlusion issues. Otherwise, I don't think the overall quality will be much better than two either. The extra coverage could reduce some cleanup work and that's all I'm really expecting. That said, the dual v2 quality really isn't bad to begin with--the point cloud data is surprisingly clean, and not nearly as noisy as v1 data.

I'll post some updated tests soon-ish. Finding time for personal projects has been tough this year but I'm finally seeing more 'me time' lately and can't wait to be working on this stuff. :)

Author:  Snapz [ Thu Nov 01, 2018 1:46 pm ]
Post subject:  Re: Setting up ps3 mover


You could probably just purchase a newer Gen 6, even Gen 5 Intel MoBo and CPU, I configured an I3 Haswell dual core CPU and a USB 3.0 mobo new, fairly cheap, and used the old case and HDD and you already bought the cheap DirectX 11 graphics card before I believe that could be used.

It ran the second Kv2 easily, no frame drops and I actually controlled it through Team Viewer, so I didn't even need another monitor, I just used a long Cat5 cable to have direct connection through the router.

Now, it has been updated to even be easier to set up distributed recording from the slave computer, or so I have read, haven't tried it :)

It's basically just a distributed recording machine that cost me $250 to put together in a couple hours, but it will also drive 6 more PS Eyes, because I added the USB 3.0 adapter card also, if I ever needed it, it is much lighter and more portable than my main machine.

Refurbished they wanted well over that and even a cheap laptop was more used, so just one suggestion for you, even if went with used components off eBay even cheaper.

I hope you can get back to recording soon, I would like to see some of your animations with even a dual Kinect v2 set up.


Author:  Greenlaw [ Thu Nov 01, 2018 1:59 pm ]
Post subject:  Re: Setting up ps3 mover

Snapz wrote:
You could probably just purchase a newer Gen 6, even Gen 5 Intel MoBo and CPU, I configured an I3 Haswell dual core CPU and a USB 3.0 mobo new, fairly cheap, and used the old case and HDD and you already bought the cheap DirectX 11 graphics card before I believe that could be used.

Thanks for the tips, Snapz!

At the moment, I actually have a GTX 960 graphics card, and fairly new SSD and power supply in the dead workstation so, you're right, a new mobo/cpu and RAM might be all I really need. I don't need the most powerful system in the world for this, just something that works with the Kinect v2.

I'll look into your suggestions this weekend.

Author:  Snapz [ Thu Nov 01, 2018 2:05 pm ]
Post subject:  Re: Setting up ps3 mover


HAHA, oh yeah, forgot about the RAM, but 4 G isn't that expensive, it's all you need, maybe even whatever DDR3 you had in the dead machine is still working.

Cool, post how it works out for you later :)


Author:  Greenlaw [ Thu Nov 01, 2018 2:13 pm ]
Post subject:  Re: Setting up ps3 mover

I just did a 'google search' and I think think you're right. Even a quick glance suggests I should be able to pick up the parts I need fairly cheaply.

I might be able to keep the RAM sticks I had...I don't think there was anything wrong with them. Before the computer died, I had trouble accessing all my RAM but I think some of the slots in the mobo were going bad because I was able to swap the chips around and still got the same lesser amount. I'll keep this in mind when choosing the new board.

Will keep you posted.

Page 6 of 12 All times are UTC - 8 hours [ DST ]
Powered by phpBB® Forum Software © phpBB Group