单幅图像的深度学习，对NYU数据集进行划分

针对分割问题，官方已经划分好了：http://cs.nyu.edu/~silberman/projects/indoor_scene_seg_sup.html 

import numpy as npimport scipy.io as siodata = sio.loadmat('splits.mat')train_idx = data['trainNdxs']test_idx = data['testNdxs']train_size = len(train_idx)test_size = len(test_idx)partitions = [0] * (train_size + test_size)for i in range(test_size):idx = test_idx[i][0]partitions[idx - 1] = 1print partitions