luisruiz's profile - activity

I am compiling 3.4.1. Yes I already check transform.hpp and it seems that I have to configure something.

GPU API Call error, invalid configuration argument when using CUDA I compiled OpenCV3 in Windows 10 with CUDA, the compi

GPU API Call error, invalid configuration argument when using CUDA I compiled OpenCV3 in Windows 10 with CUDA, the compi

Same problem here. How can you modify the Policy? I have not found a solution.

GPU API Call error, invalid configuration argument when using CUDA I compiled OpenCV3 in Windows 10 with CUDA, the compi

I updated OpenCV, from 2.4.11 to 2.4.12. I am not using gpu. I am using the following instruction to build OpenCV:

sudo cmake -D WITH_GSTREAMER=ON -D BUILD_EXAMPLES=ON -D WITH_TBB=ON -D WITH_EIGEN=ON -D CMAKE_BUILD_TYPE=RELEASE -D WITH_OPENCL=OFF -D WITH_CUDA=OFF -D BUILD_opencv_gpu=OFF -D BUILD_opencv_gpuarithm=OFF -D BUILD_opencv_gpubgsegm=OFF -D BUILD_opencv_gpucodec=OFF -D BUILD_opencv_gpufeatures2d=OFF -D BUILD_opencv_gpufilters=OFF -D BUILD_opencv_gpuimgproc=OFF -D BUILD_opencv_gpulegacy=OFF -D BUILD_opencv_gpuoptflow=OFF -D BUILD_opencv_gpustereo=OFF -D BUILD_opencv_gpuwarping=OFF -D WITH_V4L=ON -D WITH_FFMPEG=ON ..

After the update, when I try to compile a program using cmake with OpenCV 2.4.12 I get this error:

CMake Error at /usr/share/cmake-2.8/Modules/FindOpenCV.cmake:299 (message):
  The following required OpenCV components were not found: gpu

If I go back to 2.4.11 using the same cmake instruction the program compiles without errors.

Is a new function related to gpu in 2.4.12? Do I have to set another flag to get rid of it?

Thanks!! I am on the part about setting the environment variables. I got the two installations (opencv 2 and 3) working but I was wondering if there is a way to set the environment variables automatically. Right now I change them in the .bashrc file depending on which version I will be using.

I have OpenCV 2.4.11 and now I am porting some of my codes to OpenCV3, in the meanwhile I still have to do work on 2.4.11. So I have been installing and uninstalling, going back and forth between 2.4.11 and 3. Is there a way to have both installed and select which one to use? e.g. at compilation time?, that will help me a lot. What has to be done? How to select them?

I am having a similar issue but with a different code. My approach is similar:

1. Get 5 random matched points
2. Calculate the essential matrix, I am using RANSAC
3. Get R and t, four possible solutions
4. Triangulate and get 3D points
5. Select the right R and t

From what I understand R should be very close to the real one and t is good up to a scale. So in this point an optimization algorithm like bundle adjustment has to be used. The reprojection error has to be minimized. sum( d( x, PX )^2 + d( x', P'X )^2 ). Here is where I am stuck. I do not understand how to apply this efficiently since I do not see any improvement. Is it there any examples of how to do this?

I have a large matrix. I am writting data to the yaml file on each iteration. I want to know if the number format can be changed so the matrix can be represented in a more "user friendly" way and to reduce the file size.

I am writting data to a yaml file, but I noticed that when I write a matrix of floats, (or doubles) if some value is e.g. 0.01, it writes 1.0000000000000000e-02. Is there a way to set the output format to 0.01 or to shorten the output number?

Hello,

I updated Ubuntu from 13.04 to 14.04.

I am using OpenCV 2.4.9 installed from source with:

sudo cmake -D WITH_OPENNI=ON -D WITH_QT=ON -D WITH_TBB=ON -D WITH_V4L=ON -D WITH_FFMPEG=ON -D BUILD_EXAMPLES=ON -D BUILD_DOCS=ON -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local/opencv-2.4.9 ..

I am using one of the sample programs: video_dmtx.cpp. I am using it with a camera.

I can see a video but before the program starts to stream the video I get the following error:

CvCapture_OpenNI::CvCapture_OpenNI : Failed to open input file (0): Bad file extension

I think the program is trying to use the OpenNI function instead of the normal one (if there is a normal one :P). Is there a way to tell the program not to use OpenNI unless specified??? or that is the expected behavior????

EDIT 1

Also when I tried to set video properties the program crashes, I am defining the video as:

VideoCapture capture;

And then to set the properties:

capture.set(CV_CAP_PROP_FRAME_WIDTH,  640 );
capture.set(CV_CAP_PROP_FRAME_HEIGHT, 480 );
capture.set(CV_CAP_PROP_FPS,          30 );

Edit 2

I reinstalled OpenCV with

sudo cmake -D BUILD_DOCS=ON -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local/opencv-2.4.9 ..

The program still crashes when setting properties so I commented that out.

Now I have this whenever I try to capture a frame from video:

GStreamer Plugin: Embedded video playback halted; module source reported: Could not read from resource.

Edit 3

Now it seems to be working.

My PC only has USB 3.0 ports (is a GIGABYTE BRIX) which were working OK before the upgrade, now I am using the camera through a USB 2.0 hub.

Also, before the upgrade I was using OpenCV from ROS, but now in 14.04 with ROS Indigo that is not possible.

When calling a video I had to input:

/dev/video0

But now with the OpenCV installation from source I only HAVE/MUST give the number 0, why???

Thanks!! --Luis