Hi @dkurt i am creating 4-dimensional blob from image cv::dnn::blobFromImage() function but it is giving rows = -1, cols = -1 and data = " ". i am using simple code as shown below.
Mat inImage = imread("A_1.ppm", 1);
Mat rgbImage;
cvtColor(inImage, rgbImage, COLOR_BGRA2RGB );
Mat 4DImage = dnn::blobFromImage (rgbImage, 1.0, Size(rgbImage.rows, rgbImage.cols), Scalar(0, 0, 0), false, false);
return 0;
first, your cvtColor call is useless, beacuse ppm's can have at most 3 channels, and you're reading it as bgr.
then, you got rows and cols wrong in Size(rgbImage.rows, rgbImage.cols)
and last, for a multi-dimensional Mat you cannot use rows and cols, those apply only in a 2d world. use Mat::size instead:
int N = blob.size[0], C = blob.size[1], H = blob.size[2], W = blob.size[3];
there's also an operator for printing:
cout << blob.size << endl; // NO () !
@my_opencv, From the docs:
the number of rows and columns or (-1, -1) when the matrix has more than 2 dimensions
source: https://docs.opencv.org/master/d3/d63...
But Tensrflow module accepts in form of 1 x 64 x64 x 3 means in NWHC can we do something to change access?
OpenCV works in NCHW for any model. Just put the output of blobFromImage to network and enjoy. See the samples: https://github.com/opencv/opencv/tree...
Asked: 2019-04-21 02:04:56 -0600
Seen: 2,291 times
Last updated: Apr 21 '19
can you look at your broken code above again, and try to repair it ?
(so folks can try to repeat the problem ?)
Sorry @berak now i corrected.