Hi,

This is very interesting! Does it work with the latest version of Vuforia, Unity 5.3.x, and current Android / iOS?

There is a bug in UCS 1.2 which prevents the VidTexPatch example from working correctly.

The problem was caused by a new way of resetting the camera which was needed for Vuforia 2.8.7. Get the fix here:
http://pilotpage.monosock.org/fileadmin/files/UCS12bugfix.rar
Replace the two scripts in the Scripts folder.

Thanks for the update, nice!

New version uploaded: v1.2

Changelog v1.2:
-Fixed compatibility issue with Vuforia 2.8
-Simplified Vuforia installation.
-Added support for the StudierStube Tracker AR engine.
-Fixed bug which caused a crash if an image target dataset contains more markers then specified.

when this could be possible to do with vuforia Unity?

http://www.youtube.com/watch?v=T70hgNvpUxE

That can be done using PointCloud (uses SLAM technology) but you don't need UCS for that. UCS supports PointCloud though by placing the scene in the correct orientation.

Have a look here:
http://pointcloud.io/

Then download UCS and read through the manual on how to set up PointCloud and how to initialize the SLAM map. Let me know if you have any problems.

 Can u show me how to make Vuforia, SLAM tracking using PointCloud from 13th Lab!

Thanks for the update Elecman!

The installation procedure for Vuforia 1 is located in the manual. What are you not sure about?

I am trying to upgrade to Vuforia 2.0 but I am really stuck due to insufficient documentation of the API changes.

I am in the process of separating the geometry creation from the marker stitching. This will probably take a few months to complete.

The Math is present in the source code. Mainly in MarkerRecorder.cs. The best way to understand the code is to single step through it using the FrameMarkersManual scene. Then you can see how the markers are recorded and how the math works.

Let me know if you have any further questions.

Thanks for this great tool, unfortunately I could not run it with Vuforia2 yet. I appreciate if you can explain how it integrates with Vuforia1, and the math, then more people (including me) may be able to step up and upgrade it to work with newer versions.

Also it would be great if the stiching part gets separated from the geometry generation, some sort of a Universal ARCamera. I mean, One may want to take advantage of the marker hand-off mechanism without the need to extra libraries to draw and define geometries and load objects on runtime.

Any advice as to how to read your code is appreciated.

Update on the Vuforia 2.0 migration. 

Unfortunately Vuforia 2.0 breaks nearly everything on the AR part. Most of the Vuforia API changes are undocumented so I can't fix it that easily. I am waiting for better documentation and at the moment the migration project is at a halt. Sorry for the inconvenience.

Yes, the Unity forum is much better as it is way faster to load, the forum posts are not displayed in reverted order, the text formatting doesn't get messed up if you edit a post, and code tags actually work. Let's go there.

I have fixed it. Thanks for letting me know about this bug. Please download the latest version again.

Don't use the MarkerRecorder.cs script from an older version.

You can use the red car demo from an older version though. Try the link below but because the UCS version is old, ONLY extract the car and related scripts from it.

http://pilotpage.monosock.org/fileadmin/files/UCS_07.rar

Wow .. .great to hear, i thought I was missing the plot or something  ;)  

I will be setting this up over the next hours so if you are about please check my progress occassionally as I've been at this all day & now I feel I might be able to get it sorted. 

Perhaps its better I send all my messages to the Unity3D forum instead.. yes ?   

I will go to the Unity3D forum now, download the new version and set-up UCS from scratch. 

Really Hope I can get it operating this time around. 

  Mick

Hi Elecman,

Ok I have found a MarkerRecorder.cs script   ( it was not included in the set up package  v.06 )   I will try and continue ..... Are here any other such files that I should be aware of ??  

I'm confused about this line " RayCastLayerMask in the Inspector on MarkerRecorder.cs should have: everything except "ignore raycast" and vidScreen. "   

Is this referring to the MainCamera culling  or is there supposed to be a "RayCastLayerMask " prefab ?   

p.s this is my first time trying to use UCS ... so any tips would be appreciated.  

    Is the red car example available to download and test ? 

 Mick

HI Elecman,

I am following the vuforia Manual setup & in the correct order.  

 I am simply trying to add " MarkerRecorder.cs " to the execution order but  there is no MarkerRecorder.cs script  

 Can you please check to make sure it is in your latest upload or perhaps it is now re-named.  

I have check over and over through the folder etc.. but nothing.

 note: I have Vectrosity installed but not object reader or unifilebrowser.

 Regards,

              Michael

New version uploaded.

Most notable changes:
-More accurate rotational averaging when multiple markers are in view.
-Fixed compatibility issue with Vectrosity.

Quote:

Which is easier to adjust for an end user: a pre-made collision object (obj) that replicates a real world object boundary such as a table with four legs. Or vector lines as seen in your sample vid?

If the object is complex then it is easier to load a pre-made obj file. For this you need:
-Vectrosity for the vector lines. Note that in this case the user doesn't use the vectors to build the geometry. The vector lines are only used to align the obj file with the real world geometry. UCS will automatically draw a line on every obj triangle edge.
-ObjReader to load the obj files.
-UniFileBrowser to browse for the obj file you want to load.

Of course you can make your own system which replaces these software packages, since UCS is open source. For example, if you don't need the lines to be drawn, you can disregard Vectrosity. If you have your own obj loading code, you don't need ObjReader. And if you don't want to browse for the obj file but use a few pre-defined buttons instead, you can leave out UniFileBrowser.

As to what is easier to use for an end user, you have to keep in mind that you most likely don't have an obj file which can be scaled so it exactly fits say the table of the user due to the near infinite amount of geometric shapes out there. This feature is more for experimentation and research purposes then a serious commercial usage. Most users would just want to drive a model car or robot on the table and make it fall off the edge, or drive or fly something in their living room and make it bounce off the walls, floor, ceiling, etc. For that, simple rectangular planes will do. In that case, you just need Vectrosity, unless you want to write your own line code.

Quote:

What do you mean when you say, the software has been updated so it can now blend scenes? I am guessing this means if I have several game levels the object recording positions are kept for when each new scene is loaded.. yes?

The feature that you can save recordings so you can load different scenes already existed. I was actually referring to this:
Let's say that you have a milk box and you have markers attached to each side of the box. Now you can rotate the box around and have some 3d objects surrounding the box. But what if you want to have a serial box in the same frame with some markers on each side as well? In this case, the software must know that the markers belonging to the milk box are separate from the markers of the serial box. Before you couldn't do this as it was only possible to use only one milk box, one table, or whatever object. Now you can use multiple objects with a couple of markers attached to them, and move the objects around freely. This is called a merged scene. Does that make sense?

Quote:

The warp filter I see at the end of the sample video: does this allow the user to create an illusion such as floor bulging upwards? & will the car react accordingly ie. The car will drive up and over the raised section?

Yes, it does exactly that. Here is a game which uses my effect:
http://itunes.apple.com/us/app/warp-runner/id526622754?ls=1&mt=8
https://play.google.com/store/apps/details?id=com.Visionaries777.warprunner

Hi,

I am still very interested in your work and developing with UCS.

I am uncertain as to which StarScene software I require ?  

I watched your sample video several times and like the idea of vectorosity lines  but I also work alot with Maya3D obj files.

I would like to create a small car game that allows the user to choose pre built obj. shapes and scale them accordingly so they fit real world objects such as a table or chair 

So I have a couple of questions.

1.  Which is easier to adjust for an end user:   a pre-made colission object (obj) that replicates a real world object boundary such as a table with four legs.

                                             or vector lines as seen in your sample vid ?

2. What do you mean when you say,  the software has been updated so it can now blend scenes ? I am guessing this means if I have several game levels the object recording positions are kept for when each new scene is loaded.. yes ?

3. The warp filter  I see at the end of the sample video : does this allow the user to create an illusion such as  floor  bulging upwards ? & will the car react accordingly  ie.  The car will drive up and over the raised section ?

 Regards,

             Michael

Yes, it does work with IOS.

By the way, you only need UniFileBrowser if you want to load .OBJ files so you can have complex objects (stairs, table, etc) participate in collision and occlusion. You only need Vectrosity if you want to use the geometry creation feature.

Hi there

Before I go ahead and buy Vectrosity and Unifilebrowser, etc. can I just double check this all works with iOS?

Thanks

Stuart

New version added. Biggest improvement:
-Complete rewrite of the geometry creation and alignment system to make it compatible with future SLAM implementations, simplify the alignment procedure, and provide more flexibility.

EDIT:
New version added. Biggest improvement:
-Can now merge multiple scenes together, as requested.

I updated the software (v0.6 now) to make the workflow more easier to understand. The workflow is now like this:

-Place the markers around so that at least two of them can remain in view at all times.
-Record the scene and pan the camera around so all markers are into view at least once.
-During recording, select one of the markers (by tapping on it, the outline will change to bold when selected) which will be the reference marker. The reference marker is used as the world origin and for the direction of gravity.
-When the reference marker is selected and all the markers were panned into view at least once, stop the recording.
-Save the recording by selecting one of the available scenes. This scene is now automatically loaded as well. Now game mode or geometry mode can be used.

Also, there might be some problem with scene loading and saving before, but I fixed that bug in the latest version. Give that one a try and let me know if it works.

I'm getting closer to getting the UCS to work... I believe I have recorded a scene but whenever I try Game Mode or the other buttons I get the error "Load Scene first" ... when I try to Load Scene, there is an empty file dialog. After trying to save the scene, I get the prompt for selecting the reference target. I am not sure how to resolve or select the reference target?

Update released. Check out the Android App as well!

Quote:

Now I just get an empty list in UniFile for FrameMarkersRuntime

You have to record a scene and then save it before you can load it.

Quote:

White image in ImageTargetsRuntime though....

Hm, strange... Something must be wrong with with either the culling mask of the vidScreen / backgroundCamera, the background camera depth, clipping planes, or the background camera is not pointing at the vid screen. You can check all these and see if it makes sense.

Quote:

The FrameMarkersManual detects the bounding box of each frame marker but I am unable to load or save sequences.

Do not use FrameMarkersManual on a mobile device. It is to be used in the editor only. What do you mean with "unable to load or save sequences"? Do you mean a "scene" instead of "sequence"?

Quote:

Do you have an example scene from the video you showed?

The car is a modified version of the Unity Car demo. I will see if I can include it. It makes my project size much larges though...

One note: the method of operating UCS (loading, saving, etc) is included in the manual. Please read the entire manual, not just the installation instructions. Let me know if something is not clear in the manual and I will amend it.

elecman wrote:

Like this:

You cannot run ImageTargetsRuntime and FrameMarkersRuntime in the editor. On what mobile device are you running it?

Does the FrameMarkersManual scene work?

The FrameMarkersManual detects the bounding box of each frame marker but I am unable to load or save sequences.

Right now, I'm just trying to get the example running. (Do you have an example scene from the video you showed?)

I tried all three on the Galaxy Note and HTC Evo 3D. Screenshots above via ddms for Evo 3D

Okay! I had tried putting the two above Default Time... that might have been it. Before, I had QCAR before Marker, but not before Default Time.

Now I just get an empty list in UniFile for FrameMarkersRuntime

White image in ImageTargetsRuntime though....

Like this:

The actual number doesn't matter, as long as QCARBehaviour is on top. I am trying to get Qualcomm to add this as a feature so you don't have to do any mods but so far no luck.

You cannot run ImageTargetsRuntime and FrameMarkersRuntime in the editor. On what mobile device are you running it?

Does the FrameMarkersManual scene work?

-Click on any script in the Unity3d editor and select "Execution order" in the top right Inspector window. Then modify the script execution order so that QCARBehaviour.cs is executed first, and MarkerRecorder.cs second. Then click Apply.

Just to be sure, it's: QCARBehaviour, MarkerRecorder, "Default time"
and not "Default time", QcarBehaviour, MarkerRecorder
?

Still getting the same results. Also tried again after upgrading to the latest 3.5.1

I have followed the directions through several times, on different days (to be sure I wasn't too tired and misreading)

Sounds like you didn't follow the installation instructions in the UCS manual.

Trying to run the example scenes... Not sure how to get things to work...

ImageTargetsRuntime
- no camera, just white screen. buttons show up though..

FrameMarkersRuntime
- camera input is inverted?

The reason it seems that UCS only has 6 Frame Markers is because you are looking at the wrong scene. UCS has 3 scenes included: "FrameMarkersManual", "FrameMarkersRuntime", and "ImageTargetsRuntime". The FrameMarkersManual scene is only meant to be used for debugging and the Frame Markers have to be added manually I added 6 as an example, but you can add more if you like or remove them as well. The other scenes generate markers at runtime. In the MarkerRecorder.cs script there is a variable called "markerAmount". This has to be set to the maximum amount of markers you want to use. I set it to 6, but you can set it to any number as long as Vuforia supports that many markers.

You can add up to 512 Frame Markers, not 5. The amount of markers which can be tracked simultaneously is indeed limited to a much smaller number but you can still add much more markers to your scene. It just means that Vuforia can only track (5?) markers at the same time but if you pan the camera around, it will pick up others as well, as the previous ones are lost from view. You cannot fit a lot of markers in the camera field of view at the same time anyway, so this is not a problem.

With Vuforia, there is no such thing as "image markers". They are called "Frame Markers" (square with bar code around it), or "Image Targets" (picture).

UCS does support multiple data sets as Vuforia supports it, although you will have to build the user interface for that yourself. But again, the "limitation" of 5 markers is not because you can only load one data set. It doesn't work that way. One dataset can hold as much as 50 Image Targets (more but not recommended). You can only view a small amount at the same time though.

UCS supports both Image Targets and Frame Markers. In a future release you will be able to switch between those two at runtime.

So to make a long story short, UCS does not have a limitation of only 5 or 6 markers.

I notice that UCS has 6 markers but QCAR only supports tracking up to 5 image markers at the same time..

Does UCS support loading multiple data sets to allow for more than 5 image markers to be used?

Or does UCS support using both frame makers and image markers?

Yes. I didn't notice any difference.

Do square image markers work as well as rectangular ones?

The rectangle is warped because it is not aligned properly with the real world. I had to hold the camera in an awkward position to make that shot (table too big) so I was happy with it the way it looks, even though I could have done a better job.

I could have aligned it properly by applying a "Marker Plane" constraint. In the video the plane is slightly tilted, that's why the car keeps on falling off the edge and the plane also is not aligned visually.

There can be still slight inaccuracies though even if you align the geometry properly. This is caused by small marker pose errors. I can solve this by splitting the geometry up in separate triangles and fixing the vertices to a marker position or offset thereof. This is a project in development though.

Items created at runtime work fine as this is a core feature of UCS. It also works if it is scaled to be bigger then the camera frame. Any items outside of the camera frame but spanning into the current view are correctly mapped. This also is a core feature of UCS.

Also... does this work if the items in your scene are actually created at runtime?

What if one item is changed (during runtime) to be larger than the size of the camera frame, spanning to the domain of another image marker? Will the occluded part show up on the other image marker?

elecman wrote:

I added a video of UCS. Check the first post.

cool! i notice that in your video when you defined the rectangular geometry of the table using constraints, you have a warped edge on one corner. Is there a way to stabilize this so it stays a rectangle with all 90 degree angles in the plane? (See 0:18 teal rectangle and 3:39)

I added a video of UCS. Check the first post.

Update added. See first page. I fixed the different sized markers issue but it comes with special instructions due to a bug in Vuforia so be sure to read those carefully.

I was going to give it a try at the end of the week anyway ;-) thanks again for all your work and I'll let you know about the outcome!

Yeah, I think it can be done. I added it to the todo list. Or perhaps it already works as is... The tracking should work I think, but certain geometry creation constraint types will not work correctly yet. Give it a try.

Ok, thanks. I also tried on another Android device and had the same problem. It would be interesting if someone else has this problem on Android, maybe someone else could give some feedback...?
About your previous comment

Quote:

Let me check up on the image target aspect ratio again. In any case, it is important that markers with different scales are perceived to be at the same distance from the camera if that physically is true. Normally a smaller marker will be perceived to be further away. But I will have a look.

The scale that is set in Unity is represented and updated in the .xml file, therefore different scales shouldn't be a problem when using image targets in opposite to markers.
Best,
Stefan

Good to hear you got it to work.

I am not sure what the practical difference is between the two path specifiers. But here is some more info:

From Kim:
STORAGE_APPRESOURCE: same location that 1.0.6 used, these datasets must be bundled with the application
STORAGE_APP: private application storage, you can download a dataset at runtime and place it in the application directory (Internal Storage on Android)

from manual:
STORAGE_APPRESOURCE: Storage for assets bundled with the application
STORAGE_APP: Storage private to the application.

I don't have access to an Android until next month but when I do, I will fix this problem. But until that time, anyone who has an Android, is experiencing the same problem, and knows how to fix it (not using an absolute path), can jump in here.

hi elecman,
sorry for not mentioning that, i'm working on android an di managed to figure out the problem. i couldn't manage to see the resource folder of my app on the device but what i did was changing your code a bit to load the resources from the sdcard. (of course i copied the files there before...) so i really don't know what the problem is but loading the resources from the sdcard works fine now. just out of curiosity, do you know what's the difference between STORAGE_APP and STORAGE_APPRESOURCE?
best,
stefan

-Are you using Android or IOS?
-If you use IOS, are you code signing your application? If you don't the app won't find the files. See this thread here which describes the same problem:
(amongst others)
http://ar.qualcomm.at/node/2001331
-Can you get the QCAR example image targets project (beta 1.5) to work?
-Can you use a file browser on your mobile device and verify that the files are there as well, even though they are located in the unity project folder?
-The reason the recording animation does not show up is because it can't find the data set files so it does not go into recording mode. I will make an error message so the user is informed about this in the next release. If the data set is loaded successfully, the recording animation will be displayed.
-Let me check up on the image target aspect ratio again. In any case, it is important that markers with different scales are perceived to be at the same distance from the camera if that physically is true. Normally a smaller marker will be perceived to be further away. But I will have a look.

hi elecman,
thanks for your answer.
i followed all your steps and the additional hints in your answer, still without success. it seems like the app can't find the .dat file although it is in "StreamingAssets/QCAR/". the output from logcat is in the attached text file.
to the GUI problem, neither the "recording..." text, nor the red dot shows up.
for the aspect ratio, i don't see any limitation from qualcomm when it comes to a restriction for image targets. i have some apps using QCAR and they have multiple image targets with different formats. just to be sure about that.
if i can do anything else to help, please let me know.
best,
stefan