ICNR initialization for Sub-pixel convolution

一，Sub-pixel convolution can be interpreted as convolution + shuffling（卷积+洗牌），一般用于超高分辨率图像生成的上采样。

二，ICNR initialization for Sub-pixel convolution
三，ICNR initialization代码

from __future__ import division
import tensorflow as tf

class ICNR:
    """
    ICNR initializer for checkerboard artifact free sub pixel convolution    Ref:     [1] Andrew Aitken et al. Checkerboard artifact free sub-pixel convolution     https://arxiv.org/pdf/1707.02937.pdf)    Args:    initializer: initializer used for sub kernels (orthogonal, glorot uniform, etc.)    scale: scale factor of sub pixel convolution    
    """
    def __init__</