Unity support broken?

[Christophe_L_]

Hello, 

I got my R200 dev kit yesterday and would like to see the Unity demo. 

It doesn't start for me though - it freezes the app after hitting the play button in Unity. The play button never turns blue, and the whole app freezes. 

I am using Unity Personal 5.1.1f in 64bit on Windows 8.1 in 64Bit as well. 

Does anyone have any recommendations for me on this? This is my deveopment machine, and I got the latest drivers and SDK for the R200 installed.

Thanks in advance, 

C

[Christophe_L_]

Got it to work, but the results seem to be unuseable. 

Am I doing something wrong? Is everyone else seeing this as well?
 

I would appreciate screenshots from others to see if my device may be defective...

[MartyG]

I have the F200 camera so I can't test this sample myself as it's an R200 sample.  It's unlikely that the camera is defective if other R200 sample programs work okay.  It's more likely a problem with the program.  Unity changed some stuff in Unity 5.0 onward - programs written before Unity 5 get a message on start-up that the project needs to be upgraded to run in Unity 5.  

Other R200 users can probably offer insights about their own experiences with this sample.  but it's far more likely that it's an issue with running the sample in Unity 5 than that there's a problem with the camera..  

Edit: another person posted today that they were having trouble with the R200 Scene Perception sample.

https://software.intel.com/en-us/forums/realsense/topic/594240#new

[Christophe_L_]

Hello Marty, 

that other person is me as well.

[MartyG]

I was referring to the origiinator of that post, Et_Real, as the "other person".  *smiles*  Sorry I didn't notice that you'd already commented on it.

[Xusheng_L_Intel]

Chris, R200 is long range camera and best working distance is from .5 meter to 2.5 meter now. You can try the sample apps in the SDK and tell me what kinds issue you have. Thanks

Christophe L. wrote:

Got it to work, but the results seem to be unuseable. 

Am I doing something wrong? Is everyone else seeing this as well?
 

I would appreciate screenshots from others to see if my device may be defective...

[Christophe_L_]

Hello David, 

thanks for getting back to me. 

As you can see from the screenshot enclosed, the camera is getting me a depth image in which the geometry is first not closed for the person (me) in the sweet spot. Nor is the wall behind me being recognized as such and drawn correctly, it seems as if the laser light is not strong enough to identify this as a closed surface, although this wasn't further behind me than 2m. 

I get "best" results from the camera when used in low resolution for the depth image (320x240) and not when using the highest resolution of 638 by 468 pixel with 1920x1080 video. Again, see the screenshots enclosed. "Best results" for me means that the geometry of the objects in sight is closed and being recognized as one coherent mesh. Another issue is the high jittery noise in the stream.

Please: I am grateful to have been accepted to this beta and I was looking forward to this camera, yet as it stand, its results are unfortunately unuseable to us. Furthermore, other persons which expressed their interest to see its results where equally disappointed.

We were expecting depth image quality at least on par with the Kinect1. Maybe we got something wrong? What is the main purpose of the camera?

I would also be very interested in seeing some real life screenshots from other people using the camera, just to rule out the possibility that we got a defective device. We were specifically looking for a camera with a longer depth range as we have written a software to record 3d pointcloud data videos for VR and would like to support different cameras.

The form factor and the fact that it runs from a single USB port is very interesting however. It seems to me though that the camera has primarily been built for gesture control? 

[Christophe_L_]

Hello David, 

I thought about the camera's problems a long deal this weekend. I think the misunderstanding stems from the fact that Intel preferred to focus on face and gesture recognition primarily, with the depth field/image being just a side product of this. 

Most of these functions are methods offered by OpenCV (like blob and face recognition), and can be driven by the HD video image itself, and not necessarily by using the depth data, which is what we are after. 

Do I see this correctly? Maybe the F200 would then be better suitable for our needs, at least to detect the user's body. 

[Etienne_R_]

I've been battling with the unity examples and the R200 for the last couple of days. Here's what I've found out

The only samples I got working are 6 and 7. 6 pulls the video stream from the camera and displays it on a plane and 7 displays a mesh created from the point cloud the camera detects. Pretty much everything else in unity doesn't work because the samples are based on hand and gesture detection (which are features of the F200). 

You can get the color videos stream and the depth video stream, but the IR video stream doesn't seem to be supported in unity yet.

Voice recognition actions and rules work.

I couldn't get face detection and and blob detection actions and rules to trigger even though these features work in the code samples ( just not in unity).

My guess is the face detection and blob detection scripts haven't yet been updated for the R200, but who knows because the only support we have is David Lu avoiding questions and informing us of details that can be found on the product page.

[SMali10]

Hi et_real

Initially I though the IR video stream did not work in Unity, however I found that it has to be accessed a certain way in code.

For the F200 the stream is "STREAM_TYPE_IR" and for the R200 the streams are "STREAM_TYPE_LEFT" and "STREAM_TYPE_RIGHT". You have to initialize the left ir stream before accessing the right stream, otherwise it will not work.

There is a full Unity script i've posting in this thread: https://software.intel.com/en-us/forums/realsense/topic/594070 that enables the left IR camera and accesses the raw 16 bit data.

Unfortunately I can't help with the other Unity issues. I still haven't gotten an official response to my question in that thread either.

[Mike_C_3]

I'm curious if anyone has gotten up & running with Unity and the R200. To be honest, it feels like the R200 is a second-class citizen in the eyes of Intel support, and I was wondering if anyone's experience is better/different.

[DKryd]

i read somewhere that unity pro is required. not sure if that is for all samples or code.

[Eddie_Offermann]

Douglas K. wrote:

i read somewhere that unity pro is required. not sure if that is for all samples or code.

This would be true for Unity versions prior to around 5.0 - Unity Free now supports native code plugins. (Unity Free and Pro are now functionally identical with the exception that Unity Free discourages changing or eliminating the Unity splash screen)

[Eddie_Offermann]

Mike C. wrote:

I'm curious if anyone has gotten up & running with Unity and the R200. To be honest, it feels like the R200 is a second-class citizen in the eyes of Intel support, and I was wondering if anyone's experience is better/different.

I haven't had any trouble with the R200 Unity demos, myself. Collectively, we (Mirada) encountered hiccups during our production of the dance project  http://www.intel.com/content/www/us/en/technology-innovation/overview.html#/story/dancer - but those were technical hurdles mostly attributable to working with an array of multiple R200s and gaining initial comfort with the SDK and what data the sensors provided.

Just getting Windows 10, the R200, RealSense SDK, and Unity working together? No trouble.