search

xinario / SAGAN

Sharpness-aware Low Dose CT Denoising Using Conditional Generative Adversarial Network
124 stars 31 forks source link

sharpness detection #6

Closed fairynn closed 6 years ago

fairynn commented 6 years ago

I wonder why training an independent network for sharpness detection? Why not calculate the sharpness difference between synthesized images and real images and then add the sharpness loss to objective function?

xinario commented 6 years ago

Good question.

The reason is that the sharpness metric we used from the paper "lbp-based segmentation of defocus blur" is not differentiable. Therefore, in order to use gradient descent methods like Adam to optimize the loss, we resorted to a dedicated network to do that.