Questions by a Skanect beginner

[cs hui]

Hi,

I am a new user of Skanect.  I have some questions that need all members of the group to help.  My Skanect is 1.7 window 32bit version with a Kinect as scanner.  The operating system is window 10 32 bit.

1.  My current desktop has a NVIDIA GeForce 9400GT with 1GB memory, but Skanect seems not recognizing it.  Is the version of my graphic card too low that Skanect does not support it?  Can I do some setting to make it work under Skanect?  The software/driver of my graphic card is already up-to-date.

2.  I cannot make "Reconstruct->Fusion" work.  Everytime, I finished recording a scanning and proceeded to "Reconstruct->Fusion", I got an error message saying "The number of frames recorded in disk is too small.  Did you select 'All frames' recording in Prepare ->Settings->Offline recording?  'Key frames' is not compatible.  In fact, I did try the setting of 'All frames' as well as 'Key frames'  I got the same error message.  Therefore I could not do "Reconstruct=>Fusion" at all.  Is this error due to the fact that Skanect does not recognize GPU in my graphic card?

3.  I am buying a gaming laptop with a graphic card NIVIDIA GTX 960M 4GB memory.  It will run in window 10 64 bit.  Is this graphic card supported by Skanect 1.8 64bit version?

Thanks.

[Tobby Ryan]

The GTX 960M is definitely supported. I have a GTX970 with Windows 10 64bit (3gb video ram), it works great, about 30fps.   I am also using the Kinect 1.8 sdk.

Always update to the newest version of Nvidia drivers.  If you go to settings, then post a log file, the Skanect guys can help you a bit.

Unfortunately 1.7 is the last 32bit version of the product. a lot of fixes have been implemented in the 1.8,1.8.2 and 1.8.3. it is only 64bit though.

I am hoping they will soon post a few more nifty features in 2016.

Since you mention you are using a laptop. here's a suggestion. Nyko Kinect extension cable. its less than 4$, and adds 15 feet between the Kinect cable and the usb breakout. I have daisy chained 4 of them together for a total of 60 extra feet.

No I do not work for them, nor am I pushing their product, It got rid of my woes and pulling on the usb port issue to where I accidentally unplugged it once.

[cs hui]

Tobby,

Thanks for your comments.  I am still waitingfor  the help from Skanect tech. support.  From the log file, in WIN10-32bit, the program did recognize the GPU in my PC, but somehow it disables it and would not allow me to perform FUSION.  In WIn7-64bit of my laptop, I got no separate video card, of course, Skanect cannot detect the GPU.  However, it still allows me to perform FUSION. 

To me, I would like to have a WIFI 3D scanner/camera so that I need not bothering with the cable at all.  Is there any in the market and supported by Skanect?

[Tobby Ryan]

the only wireless wifi camera that is available that works with Skanect is the StructureIO device that interfaces with an ipad and transmits to Skanect, that i am aware of. I do not have one yet, but i plan to get one with tax money. (aka i need an ipad too)

[cs hui]

I finally got the answers from Skanect support team.  I want to share with the members of the group.

NVIDIA GeForce 9400GT with 1GB memory in my Desktop is not supported by Skanect, because it only supports GPU with Computing Capacity of 3.0, but the Computing Capacity of mine is below that standard.

I got my new laptop from DELL.  It comes with Window 10 and graphic card NIVIDIA GTX 960M 4GB memory.  This graphic card has Computing Capacity of 5.0 and so it is supported by Skanect.  Unfortunately, I had problem with this new machine.  When it came, I installed Kinect SDK 1.8 and Skanect into it.  Then ran Skanect but I got a blank screen.  I tried to restart the system but Window got an update.  It was a very large update, almost a completely new version of window.  Then I uninstalled Kinect SDK and Skanect.  After that, I tried to reinstall Kinect SDK.  The Installer failed to install the driver with device service error id 143.  The Kinect SDK was downloaded from MS site.  I then ran Skanect to see what happed without Kinect driver.  I got error message "Kernel security check failed" and the system automatically restarted.  Window 10 64-bit is very unstable.   In my two other Laptops, I took out Window 10 64bit and re-installed Window 7 month ago.  Now it does not support its own product! 

Window 10 32-bit in my desktop works fine. Finally I returned my DELL and now begin to search for another one.

BTW, Tobby, how is the resolution of the model you built using GTX 960M at a rate of 30 fps?  The resolution of mine is not great because it is limited at 0 to 4 fps in my desktp/laptop without any graphic card.

[Tobby Ryan]

mine is a desktop, GTX970. no mobile processor. if i need my system mobile now, i now have a gas powered generator to run outside. inside, i use a 25ft Tripp Lite isobar power strip. I have 2 50ft extensions if i have to go far to get to an outlet.

models. i have a few samples of a friend of mine at http://sketchfab.com/tobby96

The bronze statue i scanned is at: Conway Wampus Cat Statue
I still have yet to compile the base and add that part too.   ah the power of procrastination!

[Umberto Colpi]

HI Tobby!

your scan is very nice!!

I ask that you used cam ??

thank you

[cs hui]

Tobby,

Yours scans are excellent.  May I know whether you need to do some editing after your scan.  If yes, what software do you use, Meshlab, Meshmixer or others?  Thanks.

[Tobby Ryan]

Umberto, i use the Xbox 360 Kinect.sensor with the USB -Power Cable.
------------------------------------------------------------------------------------------------------------------
cs hul

Editing the scan...very good question as all models do require touch up from the way i scan.

I use Meshmixer. It's free from Autodesk and does what i need in a pinch.

Here is my workflow in a nutshell.

Skanect
-----------------
prepare/record a model.
reconstruct with a smaller bounding box if you wish to get more detail.
in the process tab, i do things manulally, i hit Fill Holes- Watertight - Low as this gives a good fill, but doesn't destroy the quality. Very low keeps the quality better but not as much fill.
I then move&crop to rotate and cut to the floor to remove some garbage.
Colorize - 1m, on, on, off
External Edit -  View mesh in Explorer
This saves as a ply file colored and mostly filled. I then move to Meshmixer for repairing.

Meshmixer
------------------
I always edit at first without looking at textures, but preserve them to edit last.
Import the ply file that was shown in View Mesh in Explorer

I use SCULPT tools, to fix issues, mainly via using the bubblesmooth/shrinksmooth/pinch tools the most
after i get the model to where all the items look like they are filled in for the most part.

EDIT
i use the MAKE SOLID tool, to close the outside shell for the most part as a single shell.
Solid Type - Automatic
COLOR TRANSFER MODE - Vertex to Vertex (to make sure all colors are not blurry or washed out)
SOLID ACCURACY - 512  (max)
MESH DENSITYy - 512 (max)
Cell Size 0 mm
offset distance 0 mm
MIN THICKNESS 10 mm
cull edges threshold 100
click on update then accept

Make sure Object browser is turned on, and you can delete the original model and keep the new one.

I then go to EDIT/Seperate shells and let it populate
this will seperate all the pockets into thier own objects.
delete the object other than shell 1. (the main object you started with. hoping this does not remove another part you need, if it does, undo and remove what you dont want)
you can then higlight the original shell you kept and hit the eyeball icon. this will make it hidden to let you see what other pockets/etc you might have.

repeat the edit/make solid/ seperate shells until the object is finished

You can then go back into SCULPT
 Select Surface instead of volume, select either the paint vertex(with appropriate color), or the smooth color brush (smooth color = photoshop healing brush...ish)
fix your texture
Export back over the edit_me.ply file

Skanect
---------------------------
in process, go to External Edit
hit Reload edited mesh from disk
this will pull in the edited version in place of the original.
you can then simplify and upload to shapeways/sculpteo/export for your own.

[cs hui]

Hi Tobby,

Thanks very much for your lesson. 

I have some more questions:-

1.  After I did the recording, reconstruction and process, I export my model into format or ply and obj.  I imported it into meshmixer for editing.  However, I found that the resolution on the screen is much lower in Meshmixer than in Skanect.  I do not know whether I should proceed to do editing or not as I am afraid that once I edit and save the model, the resolution will be reduced and the outcome will be an undistinguished object when it is printed out.

2.  When I set up with Feedback quality = high, the number of frames per second is reduced to 5 - 7 only.  If I set Feedback quality = low, the number of frames is up to 15 - 25.  What value of feedback quality should I set without affecting the quality of the model produced?

3. My CPU is quad core with speed 2.7 to 3.5GHz with memory 8GB.  My GPU is nVidia GeForce 940M with 2GB memory.  Should I use CPU or GPU in recording and reconstruction?

Best regards,

C.S. Hui

Best regards,

C.S. Hui

[skiller1234]

 Hi Hui- I'd post a quick summary of my setup as a second example.  I have the structure sensor on an air2.  Skanect is running on a laptop - with a quad i7 and GT750m - so a couple years old.  However I still am getting almost 30 fps reliably on medium (laptop plugged in of course).  When I am out I can usually get by using just a battery backup - one of those all in one jobs that has jumper cables, an air pump and a flashlight - and some AC outlets.  Anyway - I like the wireless much better than the Kinect - its just so much easier to move around and get the tight spots.  Always have to make sure to get the feet really good especially if they are going to be freestanding, hands as well as they can be hard to sculpt in...

Anyhoo - what I do - is do the capture on medium.  Then after - I go straight into "process" and colorize on high.  this gives a quick idea on how good in general the scan is - and I can show the folks pretty quickly how it turned out - always say I can make it better before finally printing it out.  

Then after the day is done - for each scan I go back and reconstruct on high. Then  - watertight low - and color high like Toby.  However then I simplify down to about 1m faces - and export as as ply.

Then I import in meshmixer for the next steps.  In meshmixer I generally 

1)  Level out, plane cut the bottom so its flat.  and save

2) Configure units to scale it - and save

3)  Meshmix (smooth, inflate, drag, select to fill holes, and extrude) to make it look better - then save.

3a)  Use the "paint vertex", and "smooth color" brushes to help with color if needed

4)  double check the number of faces and the file size - reduce it if needed....  So shapeways will take it

5)  Either add (meshmix on) a doughnut thing on top of the head so they can hang it - or "import plane" a to create a nice base under the feet.

5) then I Hollow it - the hollow tool is pretty good - and it will put the holes in it for drainage.  I often have to move the holes around to get them at the bottom of the feet, or under a wrinkle in their clothes so it can't be seen.

Note be sure to save after every step - it sucks to realize you missed something and the only recourse is to start over from the begginning.

So after all thats done I then will save again as ply - and then pull it into meshlab.

In meshlab pretty much all I do is go into the filer for "color creation and processing" and adjust the brightness and contrast which really makes it better to print - and also use the "unsharp mask" option - it really helps with clarity in the face.  

Then I save as a wrl, zip it,  - and off to shapeways.

If the scan it really jacked up - face is hosed usually,  there are things you can do.  I will sometimes go into the scan folder - and move a chunk of images out of the folder - so when you reconstruct it it only does part of the person.  like i will move out everything except the head - so I can work on it separately.  Then its easy to combine them in meshmixer.

Here are some of mine I got back from Shapeways...

skiller

[cs hui]

Hi Skiller,

Thanks very much for your instruction.  However I have some points I do not understand.

1.  I find that in Setting, if I choose Fidelity high, the GPU but not CPU is running during Capture.  I can only get 5 -- 8 fps.  If I choose Fidelity medium, then both GPU and CPU are running and I get 25 - 27 fps.  If I choose Fidelity low, CPU not GPU is running and I get 25 - 27 fps.  Please note my GPU is GTX940 with 2GB memory.  I can never get 30fps.  Why?

2.  In Process, where is high/low in Colorize?  You can turn on/off Inpaint Colorless only.  In Watertight, of course you can set low for smoothing.

3.  What is this shapeways? Is this a place where you can 3D models?  What is its address?  Can it print 3D in color?

4. I am still learning Meshmixer and Meshlab.  May I ask you questions about these two softwares later?

I used Skanect to produce a ply file.  There are two big holes at the top of the head and under the chin as usual.  After running watertight in Process, the two holes are still there. I import it into Meshmixer and run Analyzer, no hole is found, though I can still see the holes.

I do not know what tool to fix it.

Best regards

CS Hui

[Richard Grenfell]

There are two big holes at the top of the head and under the chin as usual.  After running watertight in Process, the two holes are still there. I import it into Meshmixer and run Analyzer, no hole is found, though I can still see the holes.

I do not know what tool to fix it.

 What you will find that Skanect has joined the top of head and chin holes together. Mesh Mixmeshers "analyser" tool checks for incomplete surfaces and as a result won't fix the visible holes as it doesn't know there is a problem.

In some scans I have had watertight in Skanect there has been a very large inner surface inside the scanned object. The way I go about getting filling the holes is to

-  use the select tool to remove the edge of your visible hole. if this is done correctly a single blue line should be seen around the hole.

- Using the select too double click on the blue edge line. If done correctly an orange outline will now appear around the hole.

- Now hit the "F" key and hole will be should be filled ( if the edge isn't too wonky or has too many undercuts )

 now clear the selection and work on your next hole.

If you think you have disconnected the outer surfaces from the inners surfaces you could select the entire inner surface and delete it.

I also use the same select, delete and fill technique to remove and fix the scanning of the inner arms to.

There might also be a way of performing this by making the scan solid in mesh mixer then hollowing it out again, but I haven't figure that out yet.

Cheers,

RIch

 

[Tobby Ryan]

Here is a more completed statue model of the Conway Wampus Cat Statue with Base
I finally got around to fixing the base partial scans and adding to the Wampus Cat.

[Alan Jean]

Hello Skiller1234,

Just a couple question , how high are your selfies as I have problem to get such good faces details .

.on Skanect to scan how do you  set the  OFFLINE RECORDING  with all frames or just key frames ..? I  i have it set with all Frames and another one that i wonder is the UPLINK COLOR GAIN , I set it on: lock on Record .

Last how far are you from the person when you scan...? i am around a meter away  maaybe a meter an half , i wonder if it is not to far to get the face details like lips and eyes, i set my printer for 12 cm high selfies and that maybe a bit to small to get good faces details .


I am in the process to  test it with photo instead of scanning did you try it ...?

I got a zcorp 450 for printing and it seem do the job find for other details , like belt , shoes etc...,

Thank's for any help as I am new in 3D printing ...

Chers,

Alan