python长度转换代码_类型错误：只能将长度为1的数组转换为Python标量

I am a beginner to openCV and am trying to analyse an existing code for a sudoku solver. There is this section of code which throws an error.

samples = np.float32(np.loadtxt('feature_vector_pixels.data'))

responses = np.float32(np.loadtxt('samples_pixels.data'))

model = cv2.ml.KNearest_create()

model.train(samples, responses)

The error is as follows Type Error: Only length-1 arrays can be converted to Python Scalars.

The complete traceback is as below:

C:\Study stuff\FinalProject>c:\Python27\python.exe Sudoku.py

Traceback (most recent call last):

File "Sudoku.py", line 15, in

model.train(samples, responses)

TypeError: only length-1 arrays can be converted to Python scalars

Any idea as to what the issue is?

解决方案

The error message that you are getting:

TypeError: Only length-1 arrays can be converted to Python Scalars

literally means: you have provided an array of more than one element in a place where a single value or an array of a single element was expected.

So one of the arguments passed to call model.train(samples, responses) requires and scalar... but which?

A look at the latest documentation of KNearest class, allow us to see the signature of the StatsModel.train method:

virtual bool cv::ml::StatModel::train ( InputArray samples, int layout, InputArray responses )

Apparently, a new layout argument was added. But it's meaning is a little obscure from the documentation.

Without knowledge of the contents of you file, I can't tell if you need to pass ROW_SAMPLE or COL_SAMPLE, but armed with that information, I could find a similar question, whose solution was to add cv2.ml.ROW_SAMPLE as second argument to the train method:

model.train(samples, cv2.ml.ROW_SAMPLE, responses)