Ask Your Question

Eduardo's profile - activity

2019-10-16 03:30:42 -0500 edited answer Unclear how calibrateCamera estimates stdDeviations (perhaps wrong)

Quick answer: J is the Jacobian that should be of size MxN with M the number of samples and N the number of estimated

2019-10-16 03:23:44 -0500 answered a question Unclear how calibrateCamera estimates stdDeviations (perhaps wrong)

Quick answer: J is the Jacobian that should be of size MxN with M the number of samples and N the number of estimated

2019-10-11 10:25:16 -0500 commented question Camera - robotic arm coordinate system mapping(translation+rotation)

A figure would help. You probably need 3D transformation (3D rotation + 3D translation). Look at homogeneous transform

2019-10-10 13:39:02 -0500 commented question Smarter way to set all transparent pixel to white using Java openCV

You should modify the imageData and then call once source.put(imageData ). Calling source.put(i,j,d) in the loop is not

2019-10-09 13:37:47 -0500 commented question mapping returned from cv::aruco::estimatePose

It is the second one. You can test it easily by printing the tvec values. Here is the camera frame.

2019-10-09 13:34:04 -0500 commented question Understanding tvec and rvec used in cv::drawFrameAxes

cv::drawFrameAxes() expects rvec and tvec that map from object frame to the camera frame since what this function does i

2019-10-07 04:45:15 -0500 commented question Ellipse fitting similar to ImageJ feature

I would try something like: Canny > findContours > use mouse cursor position + pointPolygonTest to get the corre

2019-10-06 07:05:07 -0500 commented question Ellipse fitting similar to ImageJ feature

I would try something like: - Canny > findContours > use mouse cursor position + pointPolygonTest to get the cor

2019-10-06 07:03:53 -0500 commented question Ellipse fitting similar to ImageJ feature

I would try something like: - Caany > findContours > use mouse cursor position + pointPolygonTest to get the cor

2019-10-02 11:12:45 -0500 commented question Position estimation on tennis court

GPS RTK could be another solution.

2019-10-01 10:06:49 -0500 edited answer Scanning QR Code from a long distance

Ability to detect a tag should depend mostly on how big it appears in the image (other criterion are image quality, imag

2019-10-01 10:05:14 -0500 answered a question Scanning QR Code from a long distance

Ability to detect a tag should depend mostly on how big it appears in the image. How big it appears in the image depend

2019-09-26 15:25:53 -0500 commented question OpenCV dnn, OpenCL and Nvidia graphic cards

There is OpenCL DNN backend in OpenCV... Show me a DNN library with OpenCL that can compete with CUDA :) For now, the

2019-09-26 15:12:00 -0500 commented question A typo in the page of 'Camera Calibration and 3D Reconstruction'

Thanks for your feedback. Yes Yes In my opinion, (Xw, Yw,Zw) and Xc,Yc,Zc would be better an non-ambiguous. Usually

2019-09-20 17:07:51 -0500 commented question BM3D strange artifacts

Maybe an issue in the OpenCV implementation. If you are lucky, the IPOL and the OpenCV implementations follow strictly

2019-09-20 17:01:30 -0500 answered a question Interpolation of transformation matrix

One solution is to interpolate directly the elements of the homography matrix, see @LBerger answer. Here a video result

2019-09-19 17:32:35 -0500 commented question BM3D strange artifacts

You should add also the original image. If you have Matlab, you can also test with the original implementation.

2019-09-19 08:53:50 -0500 commented question ORB feature detection is being confused by repetitive elements (black bars). How do I fix?

Try the Lowe ratio test like here or here.

2019-09-14 13:37:22 -0500 commented question c++ matrix optimization which is look slower than optimized python numpy ?

I am puzzled.Your vectors are 512 elements right? And 1.8 s are for 2 million of iterations doing vectors substraction o

2019-09-14 13:31:02 -0500 commented question c++ matrix optimization which is look slower than optimized python numpy ?

I am puzzled.Your vectors are 512 elements right? And 1.8 s are for 2 million of iterations doing vectors substraction o

2019-09-14 13:30:45 -0500 commented question c++ matrix optimization which is look slower than optimized python numpy ?

I am puzzled.Your vectors are 512 elements right? And 1.8 s are for 2 million of iterations doing vectors substraction o

2019-09-14 13:26:32 -0500 commented question c++ matrix optimization which is look slower than optimized python numpy ?

I am puzzled.Your vectors are 512 elements right? And 1.8 s are for 2 million of iterations doing vectors substraction o

2019-09-13 16:24:02 -0500 commented question c++ matrix optimization which is look slower than optimized python numpy ?

Crosspost here. From the other answer, what I retain is expression templates is a must to evaluate mathematical operati

2019-09-13 15:33:01 -0500 commented question c++ matrix optimization which is look slower than optimized python numpy ?

Check that MKL is correctly detected in OpenCV CMake. I don't see it in your log. It will not help in this case but will

2019-09-13 15:31:40 -0500 commented question c++ matrix optimization which is look slower than optimized python numpy ?

Check that MKL is correctly detected in OpenCV CMake. I don't see it in your log. It will not help in this case but will

2019-09-13 09:07:11 -0500 commented question c++ matrix optimization which is look slower than optimized python numpy ?

You may want to try to install a BLAS library (OpenBLAS or Intel MKL). It speeds up basic operations like matrix-matrix

2019-09-12 14:43:09 -0500 commented question fixed points in findHomography

You have to use an asymmetric chessboard pattern like this one.

2019-09-10 06:53:00 -0500 commented question Get homography to warp image using camera pose obtained from 3D tracking

Have a look at this tutorial.

2019-09-04 15:09:57 -0500 commented question is there an unnecessary code repeat?

Maybe try to remove ocl_rotate() and CV_OCL_RUN(_dst.isUMat(), ocl_rotate(_src, _dst, rotateMode)). Try to validate it w

2019-09-04 09:34:33 -0500 commented question Co-ordinate system followed in cv2.warpAffine()

Have a look at the indexing in Numpy. In short, it is img[row, col] or img[y,x].

2019-09-04 09:33:34 -0500 commented question Co-ordinate system followed in cv2.warpAffine()

Have a look at the indexing in Numpy. In short, it is img[row, col] or img[y,x].

2019-09-01 18:42:12 -0500 commented question OpenCV 4 cmake switches for best runtime performance (Linux, C++, Python)

OpenCV DNN does not use CUDA for the moment. There is a PR for that. Use OpenBLAS or Intel MKL to accelerate the basic

2019-09-01 18:41:38 -0500 commented question OpenCV 4 cmake switches for best runtime performance (Linux, C++, Python)

OpenCV DNN does not use CUDA for the moment. There is a PR. Use OpenBLAS or Intel MKL to accelerate the basic routines

2019-08-31 05:29:12 -0500 commented question Delay-induced performance hit - context switching costs?

You should add the code and the data to reproduce the issue. Maybe a profiler (Gprof, Intel VTune, ...) could give more

2019-08-31 04:10:00 -0500 commented question Delay-induced performance hit - context switching costs?

You should add the code and the data to reproduce the issue.

2019-08-28 03:56:48 -0500 commented question Using findHomography to align scanned documents; How do I avoid obviously-bad (non-planar/3D) transforms?

I would try something like line detection to get the orientation of the sheet since you should get only in-plane rotatio

2019-08-27 13:41:02 -0500 commented question ROS Kinetic OpenCV3.2.0 Contrib and CUDA

It should be something like this: catkin_make VERBOSE=1 --cmake-args -DOpenCV_DIR=your_opencv_build_dir or catkin_make

2019-08-27 13:39:12 -0500 commented question ROS Kinetic OpenCV3.2.0 Contrib and CUDA

It should be something like this: catkin_make VERBOSE=1 -DOpenCV_DIR=your_opencv_build_dir or catkin_make VERBOSE=1 -D

2019-08-27 13:35:36 -0500 commented question ROS Kinetic OpenCV3.2.0 Contrib and CUDA

It should be something like this: catkin_make VERBOSE=1 -j4 --cmake-args -DOpenCV_DIR=your_opencv_build_dir or catkin_

2019-08-27 13:35:04 -0500 commented question ROS Kinetic OpenCV3.2.0 Contrib and CUDA

It should be something like this: catkin_make VERBOSE=1 -j4 --cmake-args -DOpenCV_DIR=your_opencv_dir or catkin_make V

2019-08-27 13:32:53 -0500 commented question ROS Kinetic OpenCV3.2.0 Contrib and CUDA

It should be something like this: catkin_make VERBOSE=1 -j4 --cmake-args -DOpenCV_DIR=your_opencv_dir

2019-08-27 13:30:36 -0500 commented question Using findHomography to align scanned documents; How do I avoid obviously-bad (non-planar/3D) transforms?

Can you add the original images? How the scanned documents look like? If the document is fully visible, I would instead

2019-08-27 13:28:52 -0500 commented question Using findHomography to align scanned documents; How do I avoid obviously-bad (non-planar/3D) transforms?

Can you add the original images? How the scanned documents look like? If the document is fully visible, I would instead

2019-08-27 07:19:11 -0500 commented question ROS Kinetic OpenCV3.2.0 Contrib and CUDA

Remove everythin related to OpenCV in ROS. Build your own version of OpenCV. Clone the removed package that depends on

2019-08-21 17:38:38 -0500 commented question making the optimal target for SIFT

I suggest you to look at fiducial markers detection, e.g. Apriltag or ArUco.

2019-08-20 11:24:20 -0500 commented question Angle between wall and camera using pnpRansac and its precision

You can describe a plane by a normal and a point on the plane.

2019-08-20 08:48:53 -0500 commented question Angle between wall and camera using pnpRansac and its precision

My suggestion: forget about solvePnP, since you are using a depth camera everything is already in the camera frame ins

2019-08-20 08:46:35 -0500 commented question Angle between wall and camera using pnpRansac and its precision

My suggestion: - forget about solvePnP, since you are using a depth camera everything is already in the camera frame -

2019-08-20 08:09:05 -0500 commented answer Dynamically choose best instruction set for hardware

If you are doing something like cat /proc/cpuinfo to add flags like -mavx I suppose you want to build with full optimiza

2019-08-20 07:56:49 -0500 commented question Dynamically choose best instruction set for hardware

What do you mean by: OpenCV runs the hardware check and then decides to quit Look at CPU optimization dispatching.