Make camera frames be cast to float64

[antonl]

Currently, data is returned as is, in uint16 format. This can lead to bugs if you forget to convert it to float64 before performing manipulations on it.

I should figure out a way to improve this.