RealSense camera capture about infinitam HOT 13 OPEN

That looks great, thanks! I have to say it's not on top of my priority list, but I'll have a look if I find some time.

from infinitam.

I added support for librealsense in #39. Let me know what you think.

from infinitam.

#39 is merged and closes this.

from infinitam.

Hi,

I am trying to use this library with Intel RealSense camera. I was able to build it on my Mac OSX after several attempts but now when I run " ./InfiniTAM" form my build directory I get following errors:

usage: ./InfiniTAM [ [] ]
: path to a file containing intrinsic calibration parameters
: either one argument to specify OpenNI device ID
or two arguments specifying rgb and depth file masks

examples:
./InfiniTAM ./Files/Teddy/calib.txt ./Files/Teddy/Frames/%04i.ppm ./Files/Teddy/Frames/%04i.pgm
./InfiniTAM ./Files/Teddy/calib.txt
initialising ...
using calibration file:
trying OpenNI device:
OpenNI: Initialization ...

OpenNI: Device open failed!
DeviceOpen using default: no devices found
trying UVC device
unsupported descriptor subtype: 13
unsupported descriptor subtype: 13
unsupported descriptor subtype: 13
unsupported descriptor subtype: 13
unsupported descriptor subtype: 13
unsupported descriptor subtype: 13
unsupported descriptor subtype: 13
unsupported descriptor subtype: 13
.../git/InfiniTAM-master/InfiniTAM/ITMLib/Engine/../Utils/../Objects/../../ORUtils/MemoryBlock.h(217) : cudaSafeCall() Runtime API error : out of memory.

I really appreciate your help.

from infinitam.

Using cmake, try to disable libuvc and make sure you are using librealsense. Probably in the long term, libuvc support will be removed...

from infinitam.

Hi there,

Thanks for your reply.
I commented out the line "find_package(libuvc QUIET)" in the CMakeLists.txt and compiled the library again, here is the output I get:

using calibration file:
trying OpenNI device:
OpenNI: Initialization ...

OpenNI: Device open failed!
DeviceOpen using default: no devices found
trying UVC device
compiled without LibUVC support
trying RealSense device
There are 1 connected RealSense devices.
InfiniTAM(11038,0x7fff71718000) malloc: *** error for object 0x7ffe02416838: incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug
Abort trap: 6

I appreciate your help.

from infinitam.

Have you tried whether any of the librealsense example programs work on your system? I've not observed such an error before and can only guess, what is be going on here. Maybe for some reason librealsense fails and reports an image size of 0x0, which might then upset the malloc() in InfiniTAM.

from infinitam.

I have tried running librealsense xcode examples on my system. They usually work but sometimes they crash with this error when starting the device:

malloc: *** error for object 0x100619078: incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug

Maybe @ddiakopoulos can help here

from infinitam.

Could you by any chance have multiple versions of librealsense installed in different places? I'd expect such errors if you use the header files from one version and the binaries from a different one, and if the layout of some struct or class has changed in the meantime.

In any case, this does not seem to be a problem with InfiniTAM, which as far as I can tell works well with librealsense by now!

from infinitam.

Ok, the memory issue with librealsense is fixed now.
This is the new error I am getting :

initialising ...
using calibration file:
trying OpenNI device:
OpenNI: Initialization ...

OpenNI: Device open failed!
DeviceOpen using default: no devices found
trying UVC device
compiled without LibUVC support
trying RealSense device
There are 1 connected RealSense devices.
/Users/neda/Dropbox/git/InfiniTAM-master/InfiniTAM/ITMLib/Engine/../Utils/../Objects/../../ORUtils/MemoryBlock.h(217) : cudaSafeCall() Runtime API error : out of memory.

Thanks for your help.

from infinitam.

Hi,

I have tried using the following intel realsense cameras (R200, SR300) and only the R200 works with your code. The depth image in the gui for the SR300 looks to be erroneous as it is mostly black with only a few sparse depth pixels showing up. I have tested the camera independently and it works with the librealsense library fine. What do you think is not working properly for the SR300 to be failing but the R200 is working fine? Could it possibly be the depth_scale (range) difference in the cameras. I have been using the SR300 with the pcl_kinfu package and the depth scale is approximately 0.12 which means when the scale is applied to the 16bit depth image data is usually lost when transfering to the 16bit image type of the kinfu code (similar to Infinitam using ITMShortImage). Happy to look into it further if you can point me in the right direction?

Kind regards,

Chris

from infinitam.

Unfortunately I don't happen to have a SR300 around so can't help you very much. However, regarding the depth scale, just a week ago I've added a line to the InfiniTAM RealSense reader that should get the depth scale from librealsense, see f4a1d9d . Maybe that's at least a place to start?

from infinitam.

SR300 seems to be working fine on Linux. If anyone has calibration data it would help me a lot though.

from infinitam.