问题
I am trying to reshape a handwritten character dataset into 3D form so that it can be concatenated with digit recognition dataset. I tried multiple times, but I couldnt figure out how it can be done.
The actual digit recognition dataset has the shape (60000, 28, 28) The character recognition dataset has the shape (372450, 785) and the first column is target variable. Since excluding first column 28*28=784 there is a possibility that it can be converted to 3D same as digit dataset. Please advice on how this can be done?
I need a shape like (372450,28,28) for the entire dataframe
Thank you in advance
回答1:
An array of shape (372450, 785) cannot be made into (372450,28,28) because 28*28 is 784 not 785. But if you meant making a (372450, 784) into (372450,28,28), you could do
arr = df.column_name.values
will give a numpy array of values from the column_name field of the data frame df.
Now you can use reshape() like
arr = arr.reshape(-1,28,28)
Now arr will be of shape (372450,28,28).
回答2:
If you have a handwritten character dataset with gray scaled character, you may use following to convert your dataset into 3D,
data = data.values.reshape(-1,28,28,1)
assuming you have data as pandas dataframe.
Last argument is for channels, which is 1 for gray scaled data. If you have RGB images, there are 3 channels and thus we would have converted data into 28x28x3 matrices.
来源：https://stackoverflow.com/questions/50711786/reshaping-a-dataframe-in-python-into-3d