Issues while testing on a new image

[anjali-chadha]

I am trying to run the dehazing model on the provided sample image 'data/girls.jpg'.
Got the following error at dehazing step

>> dehaze = run_cnn(haze);
Error using convConst
A must be a 4x4 or bigger 2D or 3D float array.

Error in convMax (line 52)
        convConst('convMax',convConst('convMax',J(:,:,z),ry,1)',rx,1)';
        end

Error in run_cnn (line 29)
F3=convMax(single(F2), 3);

[xiaofeng94]

Hi, @anjali-chadha
I have met the same problem when just running the default demo.m, have you solved it?

[xiaofeng94]

I solved this problem by making the following modifications.

1. change

int *ns;
ns = (int*) mxGetDimensions(prhs[1]);

to

const size_t *ns = mxGetDimensions(prhs[1]);

2. change

int ms[3]
...
mxSetDimensions(plhs[0],(mwSize*)ms,nDims);

to

size_t ms[3]
...
mxSetDimensions(plhs[0],ms,nDims);

It seems that casting size_t* (or mwSize*) to int* is not proper in x64 PC.

[Malfurion-Learns-IELTS]

I am trying to run the dehazing model on the provided sample image 'data/girls.jpg'.
Got the following error at dehazing step

>> dehaze = run_cnn(haze);
Error using convConst
A must be a 4x4 or bigger 2D or 3D float array.

Error in convMax (line 52)
        convConst('convMax',convConst('convMax',J(:,:,z),ry,1)',rx,1)';
        end

Error in run_cnn (line 29)
F3=convMax(single(F2), 3);

Did you fix this problem?

[Malfurion-Learns-IELTS]

I solved this problem by making the following modifications.

1. change

int *ns;
ns = (int*) mxGetDimensions(prhs[1]);

to

const size_t *ns = mxGetDimensions(prhs[1]);

1. change

int ms[3]
...
mxSetDimensions(plhs[0],(mwSize*)ms,nDims);

to

size_t ms[3]
...
mxSetDimensions(plhs[0],ms,nDims);

It seems that casting size_t* (or mwSize*) to int* is not proper in x64 PC.

I have used your method. But it doesn't work. Can I send you an email to ask your guide?

[caibolun]

I solved this problem by making the following modifications.

1. change

int *ns;
ns = (int*) mxGetDimensions(prhs[1]);

to

const size_t *ns = mxGetDimensions(prhs[1]);

1. change

int ms[3]
...
mxSetDimensions(plhs[0],(mwSize*)ms,nDims);

to

size_t ms[3]
...
mxSetDimensions(plhs[0],ms,nDims);

It seems that casting size_t* (or mwSize*) to int* is not proper in x64 PC.

I have used your method. But it doesn't work. Can I send you an email to ask your guide?

You can use F3=ordfilt2(F2, 49, ones(7,7)); to replace F3=convMax(single(F2), 3);. Here, convMax is used to speedup Max pooling.

[Leslli]

I am trying to run the dehazing model on the provided sample image 'data/girls.jpg'.
Got the following error at dehazing step

>> dehaze = run_cnn(haze);
Error using convConst
A must be a 4x4 or bigger 2D or 3D float array.

Error in convMax (line 52)
        convConst('convMax',convConst('convMax',J(:,:,z),ry,1)',rx,1)';
        end

Error in run_cnn (line 29)
F3=convMax(single(F2), 3);

Did you fix this problem?
You could change the parameters like this.

void mexFunction( int nlhs, mxArray* plhs[], int nrhs, mxArray* prhs[] ) {
  int nDims, d, m, r, s; float *A, *B, p;
  const mwSize *ns = mxGetDimensions(prhs[1]);
  mwSize ms[3];

[Reiko-9k]

我通过以下修改解决了这个问题。

1. 改变

int *ns;
ns = (int*) mxGetDimensions(prhs[1]);

自

const size_t *ns = mxGetDimensions(prhs[1]);

1. 改变

int ms[3]
...
mxSetDimensions(plhs[0],(mwSize*)ms,nDims);

自

size_t ms[3]
...
mxSetDimensions(plhs[0],ms,nDims);

看来在 x64 PC 中将size_t*（或 mwSize*） 投到 int* 是不合适的。

我用了你的方法。但它不起作用。我可以给你发一封电子邮件，问你的导游吗？

Excuse me, have you solved this problem？

[SchulerTsang]

I solved this problem by making the following modifications.

1. change

int *ns;
ns = (int*) mxGetDimensions(prhs[1]);

to

const size_t *ns = mxGetDimensions(prhs[1]);

1. change

int ms[3]
...
mxSetDimensions(plhs[0],(mwSize*)ms,nDims);

to

size_t ms[3]
...
mxSetDimensions(plhs[0],ms,nDims);

It seems that casting size_t* (or mwSize*) to int* is not proper in x64 PC.

I have used your method. But it doesn't work. Can I send you an email to ask your guide?

You can use F3=ordfilt2(F2, 49, ones(7,7)); to replace F3=convMax(single(F2), 3);. Here, convMax is used to speedup Max pooling.

I did the the replacement opreration as your instrution,Howerover, another error occured.Below are the error details:
错误使用 convConst
A must be a 4x4 or bigger 2D or 3D float array.

出错 convMax (line 52)
convConst('convMax',convConst('convMax',J(:,:,z),ry,1)',rx,1)';
end

出错 run_cnn (line 29)
F3=convMax(single(F2), 3);

dehaze = run_cnn(haze);
错误使用 ordfilt2
第 1 个输入, A, 应为 二维。

出错 ordfilt2>ParseInputs (line 151)
validateattributes(A, ...

出错 ordfilt2 (line 51)
[A,order,domain,s,padopt,~] = ParseInputs(args{:});

出错 run_cnn (line 30)
F3=ordfilt2(F2, 49, ones(7,7));
any suggestion?

[FullAutoCapitalism]

I solved this problem by making the following modifications.

1. change

int *ns;
ns = (int*) mxGetDimensions(prhs[1]);

to

const size_t *ns = mxGetDimensions(prhs[1]);

1. change

int ms[3]
...
mxSetDimensions(plhs[0],(mwSize*)ms,nDims);

to

size_t ms[3]
...
mxSetDimensions(plhs[0],ms,nDims);

It seems that casting size_t* (or mwSize*) to int* is not proper in x64 PC.

Birlliant

[kzq666666]

I tried the solution above, but they don't work, Can someone give me an answer？Thanks very much!

[jsbhxmj]

错误使用 convConst
A must be a 4x4 or bigger 2D or 3D float array.

出错 convMax (第 52 行)
convConst('convMax',convConst('convMax',J(:,:,z),ry,1)',rx,1)'; end

出错 run_cnn (第 29 行)
F3=convMax(single(F2), 3);

在MATLAB上运行遇到如上问题，求助应该如何解决

[Yshelgi]

I solved this problem by making the following modifications.

1. change

int *ns;
ns = (int*) mxGetDimensions(prhs[1]);

to

const size_t *ns = mxGetDimensions(prhs[1]);

2. change

int ms[3]
...
mxSetDimensions(plhs[0],(mwSize*)ms,nDims);

to

size_t ms[3]
...
mxSetDimensions(plhs[0],ms,nDims);

It seems that casting size_t* (or mwSize*) to int* is not proper in x64 PC.

Thanks,that's useful

[Chengyunlai]

I tried these solutions, it seems to have no effect, plus the author said that the error will not report this "" A must be a 4x4 or bigger 2D or 3D float array." but will report the number of parameters error

[TuJiaX]

@SchulerTsang If you meet error like:"错误使用 ordfilt2 第 1 个输入, A, 应为二维"
you should not directly use F3=ordfilt2(single(F2),49, ones(7,7));
instead, you should operate ordfilt2 to every pages of F2, then concatenate all the results.
for i=1:size(F2,3) F3(:,:,iii)=ordfilt2(single(F2(:,:,iii)),49, ones(7,7)); end

[VoidVampire]

Thanks for the fix @xiaofeng94, I tried it on MATLAB Online and it's working.