transfer matlab of face recognition to MFC, at the time of training, use of opencv CvANN_MLP class generates a neural network, but no matter what is the input image matrix, the recognition results every time,

Watched running data breakpoint debugging, MFC generated in the parameters such as matrix and matlab are the same (except the middle part of characteristic vector and matlab symbolic opposite), including the recognition of the input matrix are all the same, is the result is wrong, so it must be opencv generated by the neural network has a problem, but still can't find the problem, the Internet also checked CvANN_MLP kind usage, feel no wrong, is there any great god comparison of the neural network in the opencv use?? Please teach

CodePudding user response:

Baidu relevant keywords,

CodePudding user response:

This, I think, you put some core code, and debug easy to analysis the result back

CodePudding user response:

Final results from the second parameter matrix to predict function, rather than directly and predict the return value of a, as a result, you try it again