Skanect 1.6.0 beta 3

[Skanect]

Dear Skanect users,

We're pleased to announce the release of Skanect 1.6.0 beta3!

You can download it here, for Windows 64 bits:

http://skanect.com/files/skanect-1.6/skanect-1.6.0beta3-win64.exe

and here, for Mac OS X (10.8 & 10.9):

http://skanect.com/files/skanect-1.6/skanect-1.6.0beta3-macosx-x64.zip

It solves a number of issues and adds some new features, among which:

- Manual color camera gain control

- USB fixes on Mac OS X

- Smaller binary size and memory footprint

Also, NVidia GPU fusion was upgraded to take advantage of the latest CUDA 5.5.

Windows users, please make sure to update your NVidia drivers to the latest version:

http://www.nvidia.com/drivers

Mac users, you will need to download and install this driver:

http://www.nvidia.com/object/macosx-cuda-5.5.25-driver.html

Please keep in mind that this version is just a beta, so we welcome your feedback and bug reports.

Also, there is no windows 32 bit version this time, sorry.

Thank you again for all the great support and feedback!

Happy scanning!

--

The Skanect team.

PS: The Structure sensor is now available on pre-order: http://structure.io/preorder. Make sure you reserve yours! :-)

[Voudas Toast-eater]

trying it as we speak, ty for the "very low smoothing" option!

[ra...@gmail.it]

sensor not detected !
why ?????

no problems with 1.6 beta 2

I have istalled 1.6 beta 3 and skanect give me a message `sensor not avaiable !`

I have i7 with gfx 680 and last drivers

Thank you

Raffaele

[ra...@gmail.it]

INFO: No support for Kinect for Windows SDK.
INFO: OpenNI2: Number of devices: 0
INFO: No support for softkinetic, skipping.

[Nicolas Tisserand]

Hi Raffaele,

What's your exact sensor model?

Thanks for the report, and sorry for the inconvenience.

> --
> You received this message because you are subscribed to the Google Groups
> "Skanect" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to skanect+u...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.

[Nicolas Tisserand]

Whooops... :-/

Due to recent changes in our windows release build machine, Kinect
sensor support was accidentally stripped out from the windows binary
of this beta.

We deeply apologize for this, I have just rebuilt and re-uploaded a
correct installer.

Skanect+Kinect users on windows, please re-download the fixed version here:

http://skanect.com/files/skanect-1.6/skanect-1.6.0beta3-win64.exe

Raffaele, this surely was the problem you reported, please re-download
and let us know.

Apologies, again.
--
Nicolas

[ra...@gmail.it]

perfect !...now my kinect for windows work fine.....

...but there is one big difference with the beta2 :

the beta 2 is faster then beta 3 in the scanning mode

I am using low mode scan and after i reconstruct in high mode.

when I scan, the fps is very unstable.....it change from 15fps to 30 when I change the scanner angulation

this not in the beta2

i suppose you have changed the scanning algorithm

the final scanning result with beta3  is a little fine then beta2

ps) suggest: If I connect  my camera sony to scanner physically with a bracket , you can develope an algorithm for integrate my photos for high res mapping materials ?
for example : when I scanning an object I can click the photos with flash.
skanect retrive the overlight flashed frame in the kinekt camera video and load a photo form my sony or other cams and load the new high res photo syncronized.
skanect will rebuild the textures with a new pictures.


thank you

[Vadas Enterprises]

Hi, Nicolas. First off, good work! :) Secondly, it works perfectly (so far) with my Carmine 1.09. Only problem is however, that I get a whopping 7 FPS on a 2 TB hard drive, 8 GB RAM, GeForce GT 520 graphics card on a Win 7 (x64) system.

If something could be done about that within the next few days, that would be GREAT!

Thank You, and have a wonderful evening! :) 

[Nicolas Tisserand]

Hi Raffaele,

I compared low-resolution CPU & GPU scanning in both beta 2 and beta 3
and cannot reproduce the performance difference you are seeing.

The FPS variability in CPU mode is normal. Were you using
GPU-acceleration before?

Regarding augmenting scans with high-resolutions color camera shots,
that's a good feature request, it's been asked a few times already,
but we don't have it planned for the near future unfortunately.

Please keep us posted about your performance issues.

Best,
Nicolas

[Nicolas Tisserand]

Hi,

What were the FPS you were seeing before with beta 2? (you can
re-download it from here:
http://skanect.manctl.com/files/skanect-1.6/skanect-1.6.0beta2-win64.exe)

Are you using GPU or CPU fusion? In what resolution?

Thanks for your help!

[Vadas Enterprises]

Hi Nicolas. In beta2, for some reason, I couldn't get it to open because it crashed. In the last stable version however, I was receiving anywhere from 1 - 4 FPS with the same computer stats as above, on GPU fusion. 

[Nicolas Tisserand]

Wait... now, I'm lost. IIUC, you're now getting better FPS than
before, right? So what's your issue, exactly?

[Vadas Enterprises]

Hi, Nicolas. My issue is that it's not getting around 30 FPS. Which I need to be able to do on your typical, overpowered computer. (Which I'm on right now). Other than that, it seems to work FLAWLESSLY.  Although, I DO have one request. Can you PLEASE add colored PLY support where the PLY is an actual point cloud, as opposed to a 3D model so I can reconstruct it easier in MeshLab?

Thank You! :) 

[Jan Krummrey]

It doesn't freeze anymore! Great. 
Now it crashes though during reconstruction. Have attached a report with this post.

Keep up the good work.

What are those calls to herokuapp? Those outbound connections make me a little uneasy...

[Nicolas Burrus]

Hi,

Does it always crash? Or only sporadically? This is happening in the
Kinect driver, which is bad news, because we have little control over
it :(

The outbounds connections are for the license check.

Thanks,
Nicolas

[Vadas Enterprises]

[K- O -S]

If I purchase Skanect 1.5 will the Beta 1.6 have unrestricted exports as well etc?

Thanks

[Nicolas Burrus]

Yes, you'll get one year of free updates, including the beta.

[IO3D]

Dear all, 

           It would be great if you can add the feature of live texture reconstruction during the scan.

Many Thanks,

                    Luca.

[James Allan]

Personally, I like the "reveal" you have with the customer when you post process the colour. They see themselves as sculpture first then you reveal the colour :) Gives people a choice as some prefer the "marble" look.

So long as you could turn it on or off as needed.

[IO3D]

Yes, I am totally with you about the WOW factor for the colourization after acquistion. However, it is more convenient to have a live texture reconstruction during the scan,in order to have an idea on which area you should focus more. Also, some clothes patterns are much harder to scan properly and it is very helpful to have a colourization during the scan and not scan twice the same person.

[Vadas Enterprises]

I agree! We NEED this BEFORE the new release, Nicolas! 

[Vadas Enterprises]

On Thursday, December 19, 2013 2:11:09 AM UTC-6, IO3D wrote:

I agree! We NEED this before the new release, Nicolas! 

[James Allan]

As you say, it depends on your focus and I haven't found having a texture overlay necessary or desirable up to this point. Ensuring good geometry is more important to me, so that's my bias :) Fixing the texture afterwards presents no problems for me and if you have good geometry, you tend to get good texture results. Not always, but generally.

Again, a good feature to have for those who need it but not high on my list of things. Better cropping tools, etc would make me happier :)

[IO3D]

Yes, you are right. It totally depends by your specific focus and application. In my case, it is very important both to colorize effectively the texture and to obtain high quality mesh. However, I tested the beta and it is very improved compared to the actual official version of the software. Good Job!

[Vadas Enterprises]

I have to half agree, and half disagree. I think that both should be in there, however, I do think that texture quality takes the trophy over mesh quality unless you're doing 3D printing, and I'm pretty sure ALOT more people can't afford a 3D printer then can. 

[Ed B]

I use my scans for 3D printing (SolidDoodle3) and 2.5D CNC carving. 

Basically I think that the 3D detail and accuracy of a single colour 3D scan is more useful for actually producing something then the colour is, but that is for people who print or carve.  The current colouring is good, and colour printing is very expensive.  It depends on your use of the product.  I like detail in a 3D scan.

The price of 3D printers are not that much and have reached the $100 mark with the Peachy Printer.(I just pre-ordered a second one.)

This product is great and just keeps getting better as I can see from these exchanges. 

If I had a wish list it would be better detail followed by mapping of the software controls to something like and external game control so that hackers can automate the scanning externally.

[Nicolas Burrus]

Thanks all for your feedback. It seems that it would make sense to
have an option for live color feedback, we'll see what we can do!

[IO3D]

This is a great news for me! Thanks!

[IO3D]

Dear all,

         It would be great if you can add the features of white balance and auto exposure block.

This will allows the users to limit the texture correction after the scan, increasing the productivity of your software.

Best regards,

                   Luca.

Il giorno venerdì 13 dicembre 2013 01:38:32 UTC+1, Skanect ha scritto:

Dear Skanect users,

We're pleased to announce the release of Skanect 1.6.0 beta3!

You can download it here, for Windows 64 bits:

http://skanect.com/files/skanect-1.6/skanect-1.6.0beta3-win64.exe

and here, for Mac OS X (10.8 & 10.9):

http://skanect.com/files/skanect-1.6/skanect-1.6.0beta3-macosx-x64.zip

It solves a number of issues and adds some new features, among which:

- Manual color camera gain control

- USB fixes on Mac OS X

- Smaller binary size and memory footprint

Also, NVidia GPU fusion was upgraded to take advantage of the latest CUDA 5.5.

Windows users, please make sure to update your NVidia drivers to the latest version:

http://www.nvidia.com/drivers

Mac users, you will need to download and install this driver:

http://www.nvidia.com/object/macosx-cuda-5.5.25-driver.html

Please keep in mind that this version is just a beta, so we welcome your feedback and bug reports.

Also, there is no windows 32 bit version this time, sorry.

Thank you again for all the great support and feedback!

Happy scanning!

--

The Skanect team.

PS: The Structure sensor is now available on pre-order: http://structure.io/preorder. Make sure you reserve yours! :-)

[Nicolas Burrus]

Hi,

1.6.0beta3 has such a feature, in the Record tab (bottom right). With
Kinect for Windows you can manually select the exposure, and with
OpenNI2 sensors you can lock it to the latest automatically determined
value.

Let us know if it does not work for you!
Thanks,
Nicolas

[Vadas Enterprises]

@Nicolas, when do you expect to have 1.6.0 with the live colors feature out? It is my money and other peoples money that funds you for your product, and I and other people have a right to know.

Thank You so kindly. 

[Nicolas Burrus]

Live coloring is not in our short term roadmap right now. Its priority
may change if it we get more requests, but at this stage I can't tell
if/when we'll do it.

[IO3D]

Dear Nicolas,

                   It works perfectly, thank you very much for the assistance.

[IO3D]

Dear Nicolas,

                   I would like to suggest you to add this important feature soon as possible. 

This would make your software much easier to use.

Many thanks,

                    Luca.

[Vadas Enterprises]

I agree, we want live color. 

[Vadas Enterprises]

Yes, we want live color!

[Ryan Hendrickson]

you replied with the same account...

I want speed rather than flash. In both scanning and processing of the mesh.

On Wednesday, January 1, 2014 11:10:13 AM UTC-6, Vadas Enterprises wrote:

Yes, we want live color!

[Vadas Enterprises]

2 against 1, therefore, I win.

[David Moore]

I would much prefer faster scanning, with a better final texture capture. I don't care about live color at all.

White balancing would be a welcome addition as well. That's lower than my desire for multiple scanners, faster scanning, and better final texture capture.

Being a developer, I know that a team can only work so fast and on only so many features.

Thanks,

David

[Jan Krummrey]

My No. 1 priority would be a stable software. My mac still crashes rather often

Live color as not important for me. I just want the best possible mesh from my scans.
I'd love to get rid of the "seams" and other artifacts. Any improvement on the resulting resolution would be more than welcome.

[Vadas Enterprises]

I'd be happy if my textures didn't all look like crap and blurry when I create my model with Skanect, WITH a Carmine, mind you. I too would be ok without live color, but if live color is what it takes in order for my texture resolution to be fantastic, then so be it.

[Voudas Toast-eater]

fixing the "seams" and more control over the reconstruction process to be able to tweak the parts that cause problems to the final mesh (structure wise and texture wise) would be my recomendations. the app is fast as it is and stable in all the enviroments that i have tested (not mac though).

live color is cool and helps to know early if the outcome is going to be good. it would be nice as an option. but if the outcome would be better with other ways and smarter coding of reconstruction, then this could be more efficient anf faster to develop (i guess)

[Vadas Enterprises]

It's not even THAT, it's the fact my textures all look like crap even on a GOOD graphics card... Here I thought this would be an alternative to those $10,000 programs that cost a fortune because this 1 only costs $100, which is fair, but I was wrong.. again.

[Voudas Toast-eater]

can you give a "crappy" example? as i'm quite satisfied by my results

[Vadas Enterprises]

You don't even need an example from me. Just scan & look at your texture. It's so blurry, it's unbelievable. I can't see hardly any facial features on my mesh... :(

[Konstantinos Papadopoulos]

For your needs it seems the x10 thousand dollar solutions would fit best. That includes both software and hardware. And by hardware I do not mean a capable computer but a scanning technology other than the kinect type of sensors.

There are plenty of images and videos on the net showing what this technology/software can offer so I guess you should rather investigate better before buying than make assumptions and then complain.

Still I think the Skanect currently offers the best value for money for what it does while doing it fast and easy.

[Jan Krummrey]

Am Donnerstag, 2. Januar 2014 23:42:33 UTC+1 schrieb Vadas Enterprises:

You don't even need an example from me. Just scan & look at your texture. It's so blurry, it's unbelievable. I can't see hardly any facial features on my mesh... :(

The Kinect has a crappy camera. Try to stitch hundreds of lousy images together while enhancing overall quality. Just not doable. If you need better color quality try the setup with a second camera:

https://groups.google.com/forum/#!topic/skanect/H73viTzwoLQ

[Ryan Hendrickson]

what sensor are you using?

[Vadas Enterprises]

Carmine 1.09

[James Allan]

I think we do need to see an example from you because the impression I get is that you're doing something wrong OR your expectations are unrealistic OR there's something wrong with your sensor. Show me one of your example scans to illustrate your point and then perhaps we can help.

I've seen various levels of quality from different people using the same type of sensor. In all the cases I've seen, it comes down to technique and your willingness to work at improving your process.

[eyebee]

@Vadas:

How are you scanning - is it hand held or on a rig?
Are you using any kind of stable lighting?
What resolution are you capturing at and are you reconstructing the mesh after capture?
What kind of frame rate are you getting and are you moving the sensor so it can build up geometry?
What do you consider to be a 'good' graphics card?

My own experience using 1.5 with an xbox kinect on a stable rig with roughly averaged lighting certainly produced some amazing scans which even animators have been blown away by. We've even 3d printed the models at 8cm and managed to get facial features intact.

"You don't even need an example from me. Just scan & look at your texture. It's so blurry, it's unbelievable. I can't see hardly any facial features on my mesh... :("

As James mentions hard to consider bad or good without seeing one of your scans. Upload one to sketchfab and at least there is a benchmark for this discussion.

Also, with respect, the guys here are working as hard as they can on a brilliant piece of software - anyone who has done software development knows that changes are far from instant. Polite requests and knowledge sharing will get you a long way in a community like this.

My two penneth worth is that live colour is not essential in our work flow but each to their own.

Best of luck.

Eyebee

[Shawn Frayne]

Hi Nicolas and all,

I've been testing out the Windows and Mac versions of the new beta3 1.6.0 with some colleagues here at our workshop -- generally it's awesome with regards to framerate and output quality relative to the previous versions. 

Here are my rig specs and the setup on my Mac, just to share the results so far:

Mac version of Beta3:

Brand new Macbook Air
++Specs:  11", Intel dual-core i7 1.7GHz with Turboboost up to 3.3GHz, 8GB SDRAM, Flash storage, Intel HD Graphics 5000), mounted to the back of the device
++Skanect settings Baseline: High feedback quality, CPU recording, Offline Recording just Key Frames, QVGA ON, Track Loss Detection OFF, Bounding Box 0.9m^3, Body Scene
++Using both a Primesense Carmine 1.09 Close Range and Asus Xtion Live Pro (similar or identical to the Carmine 1.082 as far as I can tell in performance and range)

When scanning a single person in fluorescent lighting with these baseline settings, for instance, I get between 10 - 20 FPS. It's takes about a minute to scan a person well and I lose tracking about 1/3 of the time (if I scan with the Offline Recording saving All Frames, my frame rate drops and I lose tracking more often).  The main glitch I've seen is the screen image jumps every second or so in a jarring way, which I haven't seen in the Windows version of the Beta3.

One other note, during scanning the CPU is running at 200-300% of it's rated speed of 1.7 GHz, using the turboboost I imagine.  I'll let everyone know if the laptop spontaneously combusts.

re: battery:  I lose about 1% juice per minute of running Skanect with the scanner plugged in.  Which is pretty remarkable, given the Macbook Air's typically 12 hour battery life in normal operation.

++Alternative Skanect settings #1: Low feedback quality, CPU recording, Offline Recording ALL Frame, QVGA ON, Track Loss Detection ON, Bounding Box 0.9m^3, Body Scene
With these settings when scanning a half-person I was able to get 16 - 22 FPS without losing tracking at all, and then could bump up the resolution in offline reconstruction at High feedback quality.  I've found this method is necessary for scanning groups of more than 2 people and kids.  Like, this:  (http://t.co/zPCrnuxvgb)

++Alternative Skanect settings #2: Medium feedback quality, CPU recording, Offline Recording ALL Frame, QVGA ON, Track Loss Detection ON, Bounding Box 0.9m^3, Body Scene
With these settings when scanning a half-person I was able to get 12 - 20 FPS without losing tracking at all.  Using a trick to keep the camera on the person's face for 3 seconds at the end of the scan bumped up facial features significantly and give a rather good result on just Medium feedback quality. 

Punchline: a top-of-the-line Macbook Air seems to work pretty well for scanning scenes and people on Medium settings with all OpenNI sensors I've tested and the new Beta3, so long as there is a 3 second pause on critical elements where the color has to be accurate (such as faces). 

Some other notes about other rigs I've tested:

Surface Pro 2 Tablet:  Kept freezing in a real bad way during scans using the Windows version of the new beta, over a weekend of repeated tests.  Returned.
Various 1-year old laptops with Intel dual-core i7 processors, 8GB RAM, and Nvidia graphic cards -- all work as well or better than the Macbook Air with the new Windows beta, but are heavy to tote around for point-and-shoot scanning sessions around town and have shorter battery life.

[Ruby Rieke]

New 1.6 beta works great on my Surface Pro 128 gb!!  I can't use the same license key as for my workstation.  How do I get a new license key?  Are we allowed 2 as both will never be running at the same time.  Most programs allow that.

[Nicolas Tisserand]

Hi Ruby,

I replied you via the support channel. This license problem is odd and
abnormal. Can you get back to us with more info?

Thanks!
--
NT

[Nicolas Tisserand]

Hi Shawn,

Thanks for the detailed report! Much appreciated!
--
NT

[David Saladin]

Hello 

I have install yesterday 1.6 beta 3 on my 1.5 version...and after installation i have a problem..when i scan my Body ..he scan all the Body...but when he begin to scan my head he loos me..with version 1.5 i not have this problem...so i cant finish my scan because the head is not complete... with 1.5 i have 30 FPS and with 1.6 i only have 10 FPS... my graphic driver are up to date...

All works with 1.5 ..with 1.6 only Problems when i make full scan.. i have uninstall skanect 1.6 and install 1.5 but now 1.5 wont work to...what is the Problem?

the quality of 1.6 are great  ...i hope you can fix that :-)  and the other updates are great to.. You MAKE GOOD work!!

Support for 2 Kinects? You Support the New Kinect 2.0 to?

[ra...@gmail.it]

hi,kinect v1 studio create a file .xed ...you know a procedure for convert in avi ? my idea is convert to avi, reprocess with a super sample resolution, reconvert in xed and acquire with kinect fusion with a fine resolution. are you a programmer ? do you know a programmer for develop this procedure ? excuse me for my bad english language,but i am italian and no write very well in english. thank you.

[Nicolas Tisserand]

Hi,

Here are your answers:

Skanect cannot currently load .xed files, unfortunately. Sorry.

Also, I'm unsure I correctly understood what you're trying to achieve,
but you must be aware that regular video file formats will not contain
the depth data stream that makes 3D scanning possible.

Yes, I'm a (very busy) programmer.

Your problem seems unclear to me. What exactly are you trying to solve?

Thanks,
NT

[Nicolas Tisserand]

Hello David,

Can you provide us with a more detailed description of your exact
issue, along with the hardware specifications of your machine, your
operating system version and sensor model?

Ideally, a quick video showing what's happening would help us
tremendously in understanding your problem!

Thanks,
NT

[ra...@gmail.it]

hi Nicolas

I have an idea.
In internet there is a routine for `interpolate a little photo with a little resolution` in vector mode rescaling.
This routines (see in youtube super resampling) combine the vectors and rescale the frame without a loss pixels.
The result for this process is a very clear resolution.
The kinect for windows have a 640x480 zdept resolution but it not use perfectlly this resolution...result is a very poor video images.

The result in normal kinect processing is very hard smothing and a poor resolution in the 3d reconstruction.

My idea is

1) record the kinect  zdept video and rgb video streams in avi format
2) reprocess the 2 videos with this routine for increase the resolution (http://splashurl.com/o7h99aw or see this http://splashurl.com/qz5geo5)
3) apply a noise video suppressor (http://splashurl.com/nshnnxq)
4) process the 2 streams video after.... with a skanect fusion 3d builder

i think, this is the best for final result.

simply solution for you : if you in skanect while recording can create a 2 avi video streams, i can reprocess this avi videos with an external programs (premiere, vegas, after effects, ecc...)
after, skanect can download this processed video streams for corrected and fine reconstruction.

thank you

Raffaele Mattei

[Nicolas Tisserand]

Thanks for the links, they look interesting.

Improving the resolution of the RGBD streams produced by the sensors
is a noble goal, but it's not easily achieved.

Also, the video from NEC seems to work thanks to a-priori knowledge
about the input images and an image dictionary, so that would limit
the applicability to a general purpose scanner.

Anyway, with some programming, you can already recompose videos from
.SKN scan folders as saved by Skanect. They contain color data as jpeg
images, and depth data compressed using the LZF algorithm (See:
http://docs.pointclouds.org/trunk/a00625.html,
https://code.google.com/p/lzfx/ and
http://oldhome.schmorp.de/marc/liblzf.html).

Last but not least, be aware that you're tackling a hard problem.

Good luck!
--
NT

[Russ R]

Been using beta 3 and both it and the last version crashes hard a little before 11,000 folders saved when reloading a scan. 

In other words, I can scan more than 11,000 frames without problem, (and reprocess or work with it) but then can't reopen or load the file later for additional work or manipulation. But less than 10,000 frames the program is fine. 

I did several scans of more than 11,000 frame and can no longer access them. 

Seems like a fixable bug, as Skanect can deal with the file before I close the program, but then can't open it later.

[Justin Lydle]

I have a question about the hidden settings for advanced users.

I can use this:

[BoundingBox] 

center-y-as-height-percent = 0.1 

to change the starting height of the box as you stated in a earlier reply. my question is do you have a list of parameters that can be controlled by the config file. what else can I change and control from the config file. 

[Nicolas Burrus]

You can also specify custom focal lengths by following these instructions:
https://groups.google.com/forum/?fromgroups#!searchin/skanect/focal_x/skanect/1IpxYHtJnSI/5Uo8narSyrkJ

No documentation for this though, you're entering advanced territories :)

[MartinJan gaasbeek]

one month after last post . whats the new about the beta ? is it almost done ?

-mj

Op woensdag 12 februari 2014 16:01:51 UTC+1 schreef Nicolas Burrus:

[Nicolas Burrus]

Yes, will be released very very soon now.

> For more options, visit https://groups.google.com/d/optout.

[jaideep manuja]

HI

I have Purchased Orbbec Astra   3d Scanner.  

The seller told me that this will work best with skanect .  So I download your software.

But I am not able to even link your software with the scanner.

After Connecting the scanner with the computer I open your software but I can't see any option to connect to the scanner. 

( I thought it should automatically select the scanner ).

Please Let me know the solution. I Prefer a reply to my mail id :- nisay...@gmail.com 

Thanks

 

[Tobby Ryan]

#1 Skanect works with Xbox360 kinect, Asus Xtion, Primesense Carmine, and structureio devices. #2, the newest version is 1.9.1

The download page you would have downloaded states what sensors it works with.

http://skanect.occipital.com/download/#sensors