I have a C++ class for which I created a C Dll to be used in our C# solution.
Now I need to send Images to the dll, but I don't know the proper way of doing this!
This is the C++ function signature :
std::vector<Prediction> Classify(const cv::Mat& img, int N = 2);
And this is how I went about it :
Currently I tried to create this wrapper method in the dll :
#ifdef CDLL2_EXPORTS
#define CDLL2_API __declspec(dllexport)
#else
#define CDLL2_API __declspec(dllimport)
#endif
#include "../classification.h"
extern "C"
{
CDLL2_API void Classify_image(unsigned char* img_pointer, unsigned int height, unsigned int width, char* out_result, int* length_of_out_result, int top_n_results = 2);
//...
}
Code in the dll:
CDLL2_API void Classify_image(unsigned char* img_pointer, unsigned int height, unsigned int width, char* out_result, int* length_of_out_result, int top_n_results)
{
auto classifier = reinterpret_cast<Classifier*>(GetHandle());
cv::Mat img = cv::Mat(height, width, CV_32FC3, (void*)img_pointer);
std::vector<Prediction> result = classifier->Classify(img, top_n_results);
//misc code...
*length_of_out_result = ss.str().length();
}
and in the C# code I wrote :
[DllImport(@"CDll2.dll", CallingConvention = CallingConvention.Cdecl, CharSet = CharSet.Ansi)]
static extern void Classify_image(byte[] img, uint height, uint width, byte[] out_result, out int out_result_length, int top_n_results = 2);
private string Classify(Bitmap img, int top_n_results)
{
byte[] result = new byte[200];
int len;
var img_byte = (byte[])(new ImageConverter()).ConvertTo(img, typeof(byte[]));
Classify_image(img_byte, (uint)img.Height, (uint)img.Width,res, out len, top_n_results);
return ASCIIEncoding.ASCII.GetString(result);
}
but whenever I try to run the code, I get Access violation error :
An unhandled exception of type 'System.AccessViolationException' occurred in Classification Using dotNet.exe
Additional information: Attempted to read or write protected memory. This is often an indication that other memory is corrupt.
and exception error says :
{"Attempted to read or write protected memory. This is often an indication that other memory is corrupt."}
Deeper investigation into the code made it clear that, I get the exception error in this function :
void Classifier::Preprocess(const cv::Mat& img, std::vector<cv::Mat>* input_channels)
{
/* Convert the input image to the input image format of the network. */
cv::Mat sample;
if (img.channels() == 3 && num_channels_ == 1)
cv::cvtColor(img, sample, cv::COLOR_BGR2GRAY);
else if (img.channels() == 4 && num_channels_ == 1)
cv::cvtColor(img, sample, cv::COLOR_BGRA2GRAY);
else if (img.channels() == 4 && num_channels_ == 3)
cv::cvtColor(img, sample, cv::COLOR_BGRA2BGR);
else if (img.channels() == 1 && num_channels_ == 3)
cv::cvtColor(img, sample, cv::COLOR_GRAY2BGR);
else
sample = img;
//resize image according to the input
cv::Mat sample_resized;
if (sample.size() != input_geometry_)
cv::resize(sample, sample_resized, input_geometry_);
else
sample_resized = sample;
cv::Mat sample_float;
if (num_channels_ == 3)
sample_resized.convertTo(sample_float, CV_32FC3);
else
sample_resized.convertTo(sample_float, CV_32FC1);
cv::Mat sample_normalized;
cv::subtract(sample_float, mean_, sample_normalized);
/* This operation will write the separate BGR planes directly to the
* input layer of the network because it is wrapped by the cv::Mat
* objects in input_channels. */
cv::split(sample_normalized, *input_channels);
CHECK(reinterpret_cast<float*>(input_channels->at(0).data)
== net_->input_blobs()[0]->cpu_data())
<< "Input channels are not wrapping the input layer of the network.";
}
The access violation occurs when it is attempted to resize the image meaning running this snippet:
//resize image according to the input
cv::Mat sample_resized;
if (sample.size() != input_geometry_)
cv::resize(sample, sample_resized, input_geometry_);
Further investigation and debugging (here) made the culprit visible!
This approach proved to be plain wrong, or at the very least buggy. Using this code, the image on the C++ side seemed to have been initialized properly, the number of channels, the height, and width all seemed fine.
But the moment you attempt to use the image, either by resizing it or even showing it using imshow(), it would crash the application and be giving an access violation exception, the very same error that happened when resizing and is posted in the question.
Looking at this answer, I changed the C# code responsible for handing the image to the dll. the new code is as follows :
//Dll import
[DllImport(@"CDll2.dll", CallingConvention = CallingConvention.Cdecl, CharSet = CharSet.Ansi)]
static extern void Classify_Image(IntPtr img, uint height, uint width, byte[] out_result, out int out_result_length, int top_n_results = 2);
//...
//main code
Bitmap img = new Bitmap(txtImagePath.Text);
BitmapData bmpData = img.LockBits(new Rectangle(0, 0, img.Width, img.Height), ImageLockMode.ReadWrite, PixelFormat.Format24bppRgb);
result = Classify_UsingImage(bmpData, 1);
img.UnlockBits(bmpData); //Remember to unlock!!!
and the C++ code in the DLL :
CDLL2_API void Classify_Image(unsigned char* img_pointer, unsigned int height, unsigned int width, char* out_result, int* length_of_out_result, int top_n_results)
{
auto classifier = reinterpret_cast<Classifier*>(GetHandle());
cv::Mat img = cv::Mat(height, width, CV_8UC3, (void*)img_pointer, Mat::AUTO_STEP);
std::vector<Prediction> result = classifier->Classify(img, top_n_results);
//...
*length_of_out_result = ss.str().length();
}
And doing so rectified all access violations I was getting previously.
Although I can now easily send images from C# to the dll, I have some issue with the current implementation of mine.
I dont know how to send the opencv type from c# to the needed function, currently I'm using a hardcoded image type as you can see, and this begs the question, what should I do when my input image is grayscale or even a png with 4 channels ?
I would be grateful if anyone could assist me in this rectifying these issues.
