python - OCR Hand written data showing error in svm.train() -

i using opencv3.1.0 python2.7 . have implemented code of ocr hand written data here.

responses = np.float32(np.repeat(np.arange(10),250)[:,np.newaxis]) svm.train(traindata,cv2.ml.row_sample, responses) 

and getting these error

svm.train(traindata,cv2.ml.row_sample, responses) cv2.error: c:\builds\master_packslaveaddon-win64-vc12-static\opencv\modules\ml\src\svm.cpp:1618: error: (-5) in case of classification problem responses must categorical; either specify vartype when creating traindata, or pass integer responses in function cv::ml::svmimpl::train

note: since working on opencv3.x have used cv2.ml.svm wherever necessary , rest same

and if using

responses = np.int32(np.repeat(np.arange(10),250)[:,np.newaxis]) 

getting 0 accuracy

try using pytesseract. better training svm classifier. if want check out, follow link. advanced example check out site well.