Thanks for the keras port of the CRF-RNN module! For my needs, I required a version that works with 3D grayscale images, aka 3 spatial dimensions and 1 channel dimension. So, I made a few modifications to support that - 2 or 3 spatial dimensions, and an arbitrary number of channel dimensions.
Thanks for the keras port of the CRF-RNN module! For my needs, I required a version that works with 3D grayscale images, aka 3 spatial dimensions and 1 channel dimension. So, I made a few modifications to support that - 2 or 3 spatial dimensions, and an arbitrary number of channel dimensions.