CamVid~num_class = 32 ,miou is low

[Water2style]

i train about 100 epochs,and the miou is abot 25%,
i use CamVid and set the num_class to 32,is this the reason ?

[Po-Chih Huang]

There are lots of reason to cause this.

1. Deep learning technique is not stable.
2. The implemented method in this repo is old, so the miou isn't competitive.
3. In reality, semantic segmentation will need lots of pre-process and post-process (such as CRF).