This repository is modified from Caffe version of [yjxiong](https://github.com/yjxiong/caffe/tree/mem) and [DeepLab v2](https://bitbucket.org/aquariusjay/deeplab-public-ver2) for evaluation. We merge the batch normalization layer named 'bn_layer' in the former one into the later one while keep the original 'batch_norm_layer' in the later one unchanged for compatibility. The difference is that 'bn_layer' contains four parameters as 'slope,bias,mean,variance' while 'batch_norm_layer' contains two parameters as 'mean,variance'. Several evaluation code is borrowed from [MIT Scene Parsing](https://github.com/CSAILVision/sceneparsing).

For installation, please follow the instructions of [Caffe](https://github.com/BVLC/caffe) and [DeepLab v2](https://bitbucket.org/aquariusjay/deeplab-public-ver2). To enable cuDNN for GPU acceleration, cuDNN v5 is needed as required in 'yjxiong'. If you meet error related with 'matio', please download and install [matio](matio) as required in 'DeepLab v2'.