Question I'm curious about- I think we should have a fixed intensity scale and datatype for the input images. I'd vote for having the inputs always float32, normalized to [0, 1]. Does this make sense to everyone? The other option would be to have inputs always be unsigned char 0-255, or to have any type and range of input and require that methods do their own normalization.
Question I'm curious about- I think we should have a fixed intensity scale and datatype for the input images. I'd vote for having the inputs always float32, normalized to [0, 1]. Does this make sense to everyone? The other option would be to have inputs always be unsigned char 0-255, or to have any type and range of input and require that methods do their own normalization.