I am reading an image of size 1600x1200 as greyscale and then trying to access pixel value at location (1201,0). I get segfault in the end as shown in comments:
Mat gray_image = imread(argv[1], CV_LOAD_IMAGE_GRAYSCALE); // Read the file
if(! gray_image.data ) // Check for invalid input
{
cout << "Could not open or find the image" << std::endl ;
return -1;
}
const int width = gray_image.cols;
const int height = gray_image.rows;
cout<<gray_image.cols<<endl; // 1600
cout<<gray_image.rows<<endl; // 1200
cout<<(int)gray_image.at<uchar>(1201,0)<<endl; //SEGFAULT
You already stated that you are reading an image of size 1600x1200, then how can you access 1201 element from a total of 1200 rows only, actually you have misunderstood the convention of mat.at<>(). I would recommend you to use mat.at<>(cv::Point(p_x, p_y)) in this way you would never get confused with rows and cols to be passed in the mat.at<>().
EDIT : However, creating a new cv::Point is not a recommended way of accessing pixels as suggested by #Micka, So consider it as a workaround and don't rely completely on cv::Point() to access a pixel. The best possible way to iterate the pixels is defined in Opencv documentation.
It should be .at(row,col) not (col,row) and you dont have a 1201th row thats why its a seg fault
Related
I am loading a pre-trained TensorFlow model in the opencv dnn module using the following code -
cv::dnn::Net net = cv::dnn::readNetFromTensorflow("frozen_inference_graph.pb",
"graph.pbtxt");
net.setPreferableBackend(cv::dnn::DNN_BACKEND_CUDA); //Run model on GPU
net.setPreferableTarget(cv::dnn::DNN_TARGET_CUDA);
Mat image = imread("img.jpeg");
Mat resized;
cv::resize(image, resized, cv::Size(300, 300));
cout<<resized.size()<<endl;
cout<<"Resized"<<endl;
auto input_image = dnn::blobFromImage(image, 1.0, cv::Size(300, 300),
cv::Scalar(127.5, 127.5, 127.5),
false, false, CV_32F);
cout<<"Now setting Input";
net.setInput(input_image);
auto detections = net.forward();
cout<<detections;
return 0;
However the I get the following error as mentioned in the question -
what(): OpenCV(4.4.0) /home/atharva/opencv-4.4.0/modules/core/src/out.cpp:87: error: (-215:Assertion failed) m.dims <= 2 in function 'FormattedImpl'
Could someone please point out what the mistake is?. I believe there is some problem in BlobFromImage as nothing after it is getting printed.
TIA
This error occurs because you are trying to print a cv::Mat to standard output that has more than 2 dimensions. With cv::dnn, the output after using net.forward() is 4-dimensional. However I have no idea what model you are using because the output structure of the blob is different depending on what task you are trying to do. If I had to guess you are doing some sort of object detection given your choice of variable names. In that case, usually the first dimension is the batch size and since you are using only one image, the batch size is 1. The second dimension is the number of channels in the output. As you're doing object detection on the image, this will also be size 1. The third and fourth dimensions are the number of rows and columns for the final output layer.
Going on faith, you can extract a 2D version of this cv::Mat to print out to standard output by doing:
cv::Mat output(detections.size[2], detections.size[3], CV_32F, detection.ptr<float>());
Now that this is a 2D matrix, you can print out this instead by doing std::cout << output << std::endl;.
I've c++ and OpenCV 3.1 and i separated the RGB three channels with these code :
Mat src = imread("original_image.jpg",CV_LOAD_IMAGE_COLOR);
Mat bgr[3] , bluemat ;
split(src,bgr);
bluemat = bgr[0];
std::cout << "bluemat.at<int>(0,1); : " << bluemat.at<int>(0,1) << '\n';
The strange thing is it print out a big number : 1415208581 , why is that ?
Isn't it suppose to be in 0-255 range ? why it is not ?
(expanding comment for search)
A feature of openCV is that the cv::Mat image type is a very simple object based on the original C version of the library.
The cv::Mat contains a field giving the pixel type but the data is stored simply as a block of memory bytes. You can use this block of memory to store and access pixels however you want. It makes it very flexible and easy to link with other libraries, but means you need to manually keep track of what the data is.
So data.at<int>(0,0) will extract the first pixel as if the data was ints (whatever int is on your platform), data.at<uchar> (0,0) will get the first pixel as a byte (which is generally what you want).
The main difference here is casting the memory byte to uchar vs int. What form the data you have depends on how you read it in. CV_LOAD_IMAGE_COLOR read the image as 16-bit/32-bit values. I cannot compile the code you gave me becuase of the memory issues that are created by converting that data to an int.
Now, if you use uchar, that will be solved. The problem however with the cout printing a character, has to do with the overload functions of <<
Mat src = imread("1.jpg", CV_LOAD_IMAGE_COLOR);
Mat bgr[3], bluemat;
split(src, bgr);
bluemat = bgr[0];
std::cout << "bluemat.at<uchar>(0,1); : " << (int)bluemat.at<uchar>(0, 1) << '\n';
What I changed was just to insert the typecast int. This issue is expanded here, but tells the output stream that this is a number, not a character.
I am having issues with reading a recording I made using the Recorder class in openni2 with the asus Xtion PRO LIVE. The problem is that once every ~50 frames a wrong frame is read from the recording, this was tested by storing the generated image (converted to an opencv matrix) as an .png image (using the opencv imwrite function) with an index. Shown below is my code, I also tried the code posted in the "OpenNI2: Reading in .oni Recording" question. This also doesnt work. The videostreams, device, videoFrameRefs and the playback controller are all initialized in an initialization function. I can post this as well if necessary.
openni::VideoStream depthStream;
openni::VideoStream colorStream;
openni::Device device;
openni::VideoFrameRef m_depthFrame;
openni::VideoFrameRef m_colorFrame;
openni::PlaybackControl *controler;
typedef struct _ImagePair {
cv::Mat GrayscaleImage;
cv::Mat DepthImage;
cv::Mat RGBImage;
float AverageDistance;
std::string FileName;
int index;
}ImagePair;
ImagePair SensorData::getNextRawImagePair(){
ImagePair result;
openni::Status rc;
std::cout<<"Getting next raw image pair...";
rc = controler->seek(depthStream, rawPairIndex);
if(!correctStatus(rc,"Seek depth"))return result;
openni::VideoStream** m_streams = new openni::VideoStream*[2];
m_streams[0] = &depthStream;
m_streams[1] = &colorStream;
bool newDepth=false,newColor=false;
while(!newDepth || !newColor){
int changedIndex;
openni::Status rc = openni::OpenNI::waitForAnyStream(m_streams, 2, &changedIndex);
if (rc != openni::STATUS_OK)
{
printf("Wait failed\n");
//return 1;
}
switch (changedIndex)
{
case 0:
rc = depthStream.readFrame(&m_depthFrame);
if(!correctStatus(rc,"Read depth")){
return result;
}
newDepth = true;
break;
case 1:
rc = colorStream.readFrame(&m_colorFrame);
if(!correctStatus(rc,"Read color")){
return result;
}
newColor = true;
break;
default:
printf("Error in wait\n");
}
}
//convert rgb to bgr cv::matrix
cv::Mat cv_image;
const openni::RGB888Pixel* colorImageBuffer = (const openni::RGB888Pixel*)m_colorFrame.getData();
cv_image.create(m_colorFrame.getHeight(), m_colorFrame.getWidth(), CV_8UC3);
memcpy( cv_image.data, colorImageBuffer,3*m_colorFrame.getHeight()*m_colorFrame.getWidth()*sizeof(uint8_t) );
//convert to BGR opencv color order
cv::cvtColor(cv_image,cv_image,cv::COLOR_RGB2BGR);
result.RGBImage = cv_image.clone();
//convert to grayscale
cv::cvtColor(cv_image,cv_image,cv::COLOR_BGR2GRAY);
result.GrayscaleImage =cv_image.clone();
//convert depth to cv::Mat INT16
const openni::DepthPixel* depthImageBuffer = (const openni::DepthPixel*)m_depthFrame.getData();
cv_image.create(m_depthFrame.getHeight(), m_depthFrame.getWidth(), CV_16UC1);
memcpy( cv_image.data, depthImageBuffer,m_depthFrame.getHeight()*m_depthFrame.getWidth()*sizeof(INT16) );
result.DepthImage = cv_image.clone();
result.index = rawPairIndex;
rawPairIndex++;
std::cout<<"done"<<std::endl;
return result;
}
I also tried it without the using waitForAnyStream part but that only made it worse. The file I am trying to load is over 1GB, not sure if that is a problem. The behaviour seems random because the indexes of the wrong images are not always the same.
UPDATE:
I changed the seek function to seek in the color stream instead of the depth stream. I also ensured that each stream is only waited for once in the waitForAnyStream function by setting the corresponding point in m_streams to null.
I found out that it is possible to find the actual frame index for each frame (.getFrameIndex) so I was able to compare the indices. After getting 1000 images there where 18 wrong color images, 17 of which had an index error of 53 and 1 had an error of 46. The depth images had an almost constant index error of 9.
The second thing I found was after adding a sleep of 1 ms (also tried 10ms but the results where the same) before the waitForAnyStream function the returned indices change significantly. The indices dont make large jumps anymore but the color images have a standard offset of 53 and the depth images have an offset of 9.
When I change the seek function to seek in the depth stream then with the delay the color stream has a constant error of 46 and the depth stream has an error of 1. Without the delay the color stream has an error of 45 and the depth stream has no error with occasional spikes of errors of 1.
I also looked at the indices when I dont use seek and waitForAnyStream but just do it as is proposed as an answer in "OpenNI2: Reading in .oni Recording". This shows that when the file is read by just calling m_readframe multiple times then the first color frame has index 59 instead of 0. After checking it turns out that frame 0 does excist and is an earlier frame than frame 59. So just opening the file and using readframe doesnt work at all. There are however no index errors, just like when I added the sleep to my own implementation.
UPDATE2:
I have come to the conclusion that either the openni2 library doesnt work properly or I have installed it incorrectly. This is because I am also having problems setting the Xtion to a resolution of 640x480 for both streams. When I do this the depth image only gets updated once every ~100 seconds. I have written a quick fix for my original problem by just filtering out the frames who's indices are wrong and just continuing with the next image.
I would still like to know how to fix this, so if you know please tell me.
UPDATE 3:
The code I am using for setting the framerate and fps for recording is:
//colorstream
openni::VideoMode color_videoMode = m_colorStream.getVideoMode();
color_videoMode.setResolution(WIDTH,HEIGHT);
color_videoMode.setFps(30);
m_colorStream.setVideoMode(color_videoMode);
//depth stream
openni::VideoMode depth_videoMode = m_depthStream.getVideoMode();
depth_videoMode.setResolution(WIDTH,HEIGHT);
depth_videoMode.setFps(30);
m_depthStream.setVideoMode(depth_videoMode);
I forgot to mention that I also tested the resolution by running the sampleviewer example program (I think, it was a few months ago) and changing the resolution in the xml config file. Here the color images would be shown fine but the depth images would update verry rarely.
I have used Canny edge detector to successfully identify the edges of a given image. I'm struggling with finding specific points on this detected edge line.
My approach:
I used the cv::canny function in opencv and the output is stored in cv::Mat format. I want to iterate through the all values of the matrix and identify all those pixels where the edge is present so that I can detect the specific points on the detected edge line.
Function used:
cv::Canny(frame_gray,contours,50,150);
The output is stored in contours and it is of type CV_8UC3
To access the pixel value, have tried
contours.at<int>(i,j) != 0
and also
contours.at<uchar>(i,j) != 0
Will greatly appreciate help in the above. If the approach is correct and am missing something or else if i should try another approach
Thanks
Edit:
for(int i=0;i<img_width;i++)
{
if((int)contours.at<uchar>(i,neckcenter.y) > 0 )
{
Point multipoints(i,neckcenter.y);
circle( contours, multipoints, neckpoint, Scalar( 255, 0, 0 ),4, 8, 0 );
cout << (int)contours.at<uchar>(i,neckcenter.y) << endl;
}
}
I am using the above code which forms a small circle of radius 1 (defined by neckpoint) where it detects a point on and edge. The neckcenter.y is a constant value derived from an earlier calculation. What am i doing wrong here ?
Output of the code -
you probably want a grayscale pass before applying Canny:
Mat gray;
cvtColor(bgr,gray,CV_BGR2GRAY); // now gray is a 8bit, uchar Mat
Mat contours;
cv::Canny(gray,contours,50,150);
// now you're safe to use:
uchar value = contours.at<uchar>(i,j);
The syntax:
contours.at<uchar>(i,j)
Is correct for your case in terms of data type (i.e. a grayscale image). The problem is possibly hinted at by this line:
for(int i=0;i<img_width;i++)
When you access OpenCV pixels using at, you must specify the pixel position as (row, col), so your indexing is the wrong way round. Try this in all places where you access pixels:
contours.at<uchar>(j,i)
From the OpenCV documentation:
You have a 3 channel image of the type unsigned char. To access it you should use the cv::Vec3b type. Here is how to do it:
int channel = 0;//or 1 or 2
contours.at<cv::Vec3b>(i,j)[channel]
To check if all elements are 0:
contours.at<cv::Vec3b>(i,j)[0]==0 && contours.at<cv::Vec3b>(i,j)[1]==0 && contours.at<cv::Vec3b>(i,j)[2]==0
But where do you have the information that the image type of contours is CV_8UC3 ?
CompVision once again, I'm working with jpeg images in my application. Just because I'm a bit familiar with MFC and ATL, I used CImage to access pixel values.
For my needs I calculate brightness matrix for the image during initialization. Function goes like this (Image is the name of my own class, unimportant, bright is float[][]):
void Image::fillBrightnessMatrix(){
COLORREF val;
for(int i=0;i<width;i++){
for(int j=0; j<height;j++){
val=src.GetPixel(i,j);
bright[i][j]=rgb_to_L(val);
}
}
}
Where src is an instance of CImage class, rgb_to_L - some function that calculates brightness of the color.
Examining the performance of my app, I discovered that GetPixel is the most expensive operation, and it significantly (really, ~700 times slower than any other operation) slows down the whole initializing of image. The question is, which library can you suggest for fast access to single pixel values? I don't need any other operations but loading jpeg image and accessing single pixels. Performance is important, because my application works with set of ~3000 images and I can't wait for hours to get results.
Use CBitmap::GetBits() to get a raw pointer to the pixel data. You can now directly party on the pixels without going through the expensive GetPixel() method. There are a number of things you need to be careful with when you do this:
You have to use CBitmap::GetPitch() to calculate the offset to the start of a line. The pitch is not the same as the width.
Lines in the bitmap are stored upside-down
You have to deal with the pixel format yourself. A 24bpp image stores 3 bytes per pixel. An indexed format like 8bpp requires looking up the color in the color table. 32bpp is the easy one, 4 bytes per pixel and the pitch is always the same as the width.
I always recommend OpenCV.
This is a humble code snippet to get you started:
IplImage* pRGBImg = cvLoadImage("c:\\test.jpg", CV_LOAD_IMAGE_UNCHANGED);
if (!pRGBImg)
{
std::cout << "!!! cvLoadImage failed !!!" << std::endl;
exit(1);
}
int width = pRGBImg->width;
int height = pRGBImg->height;
int bpp = pRGBImg->nChannels;
for (int i=0; i < width*height*bpp; i+=bpp)
{
if (!(i % (width*bpp))) // print empty line for better readability
std::cout << std::endl;
std::cout << std::dec << "R:" << (int) pRGBImg->imageData[i] <<
" G:" << (int) pRGBImg->imageData[i+1] <<
" B:" << (int) pRGBImg->imageData[i+2] << " ";
}
You should probably extract the jpeg to raw data, then access the raw data instead of GetPixel.