I wrote code sift and knn but performance is very low so ı want to use svm for classify. but ı have 10 class and svm doesn't support multi class so I wrote code multiclass svm (http://www.mathworks.com/matlabcentral/fileexchange/39352-multi-class-svm) but matlab give an eror for k means=150. this is :
Error using seqminopt>seqminoptImpl (line 198)
No convergence achieved within maximum number of iterations.
Error in seqminopt (line 81)
[alphas offset] = seqminoptImpl(data, targetLabels, ...
Error in svmtrain (line 499)
[alpha, bias] = seqminopt(training, groupIndex, ...
Error in multisvm (line 20)
models(k) = svmtrain(TrainingSet,G1vAll);
Error in den (line 57)
results = multisvm(histtrain,grupp,histogramtst);
.This error only kmeans k=>150. I wrote code for k means k=30 and it doesn't give an error.
I would suggest you using libsvm, which supports multi-class svm classification.
http://www.csie.ntu.edu.tw/~cjlin/libsvm/
For SIFT computation, I suggest you using vlfeat, which supports both grid-based dense sampling and detector-based solutions.
http://www.vlfeat.org/
Both tools are public and support matlab interface. I suppose you are using K-means for building the codebook. I am not sure why your code gave you the error message, but I believe it can be solved by using the tools suggested above.
I support Xiaomeng's suggestion. I have used LIBSVM (and LIBLINEAR) for multiclass SVM classification and the VLFeat implementation of SIFT. Highly recommended!
I suggest to use python code for the implementation of LIBSVM, multi-class classification is highly supported by the functions of LIBSVM, as constructed model will enhance the performance at large scales data-sets .
ı want use libsvm for classifying my test dataset but ı can't. I went to "File > Set Path > Add folder > [Select the LBSVM folder] > Save . And ı run this code:
SVMStruct = svmtrain(histtrain,grupp);
snc = svmclassify(SVMStruct,histogramtst)
But matlab gives this error:
SVMTRAIN only supports classification into two groups. GROUP contains 10 groups.
Can you help me ı have 405 images for train dataset and ı have 192 images for 192 images what can ı do
- Badges
- Science topic