Multi-output Regression Models?

My input is image data, and my target values are a vector of numbers in the range [0,1].

I'm wondering if there's some good techniques out there for regression on image data. And I'm looking into multi-output models, but I'm open to anything. I'm guessing the only advantage to multi-output is that it saves one from training multiple models.

I was thinking deep neural nets, because of their track record with image data. How would I perform regression with neural nets? Should they be RBMs or convolutional?

How would you tackle this task?