Fix typo

• Use float32 instead of numpy.float (they prefer to specify the size, else to use python float)
• Fix actual detector size to return python int instead of numpy data