The main difference between this model and other state-of-the-art models is the combination of
pixel-level CRFs and DCNN-based ‘unary terms’.

1. How to get dense inference form cousre FCNN output?

Hole algorithm:

We skip subsampling after the last two max-pooling layers and modify the convolutional filters in the layers that follow them by introducing zeros to increase their length (2x in the last three convolutional layers and 4x in the first fully connected layer). We can implement this more efficiently by keeping the filters intact and instead sparsely sample the feature maps on which they are applied on using an input stride of 2 or 4 pixels, respectively.