time: 18.5709ms -> ~53.8478Hzīut for some reason, nothing is being published on any of the topics. Failed to decompress rgb image! TurboJPEG error: 'Not a JPEG file: starts with 0x3b 0xdd' found device: GeForce GTX 780 Ti vendor: NVIDIA Corporation But I haven't found any nvidia-openclĭriver that provide opencl >1.2 to my GPU.Īlso pitifully the CPU is old Xeon without Core-gpu.
#Cuda driver version 6.5.18 upgrade#
Perhaps I need to upgrade my ubuntu distribution. Topic depth: /kinect2_head/depth_lowres/imageĪlso I tried rviz, as I subscribed to the image topic, and resulted in just black image. Topic color: /kinect2_head/rgb_lowres/image Rosrun registration_viewer viewer -kinect2 -image The output is the same as Protonect, but as I run: found device: Quadro 4000 vendor: NVIDIA Corporation found platform: NVIDIA CUDA vendor: NVIDIA Corporation The output differs, If using opencl, u got output like: (no such output for CPU) To check your GPU compute capability, see the ComputeCapability property in the output of the gpuDeviceTable and gpuDevice functions. The cc numbers show the compute capability of the GPU architecture.
#Cuda driver version 6.5.18 registration#
When the depth registration is initialized in line 143 of the kinect2_bridge you can choose CPU instead of OPENCL. To see support for NVIDIA GPU architectures by MATLAB release, consult the following table. For the IAI_Kinect2 ros package, I was trying to compile using CPU by:Ĭomment out "#define CV_OCL" in line 31 of the kinect2_bridge and it will not use any opencv opencl functionality at all. Since repos in 12.04 is opencl1.1, while I have tried opencl1.2, which may cause link error, which requests ocl-icd-opencl-dev package which no repos for 12.04 and also get conflict with nvidia-opencl. Qz1: Is that the version of the opencl that make it so? Should I use a opencl version >1.2? Though a little bit lag(Soon I figure out the GPU is not available, almost costing 1 cpu core)ĬL_PLATFORM_VERSION: OpenCL 1.1 CUDA 6.5.18ġ devices found supporting OpenCL on: NVIDIA CUDA I could compile the libfreenect2 and by running the Protonect and all the data (RGB depth ir ) could be visualized. I am trying to deal with the IAI_Kinect2 package in ubuntu 12.04 ros- hydro, I wonder if any one got the correct data? So this is what I got:
0 kommentar(er)
