Extra-feature: read float images.
This caffe version (https://github.com/BVLC/caffe) supports the following architectures. 0. Deconvolution Network: Learning Deconvolution Network for Semantic Segmentation Hyeonwoo Noh, Seunghoon Hong, Bohyung Han [http://arxiv.org/abs/1505.04366] 0. Decoupled Network: Decoupled Deep Neural Network for Semi-supervised Semantic Segmentation Seunghoon Hong, Hyeonwoo Noh, Bohyung Han [http://arxiv.org/abs/1506.04924]