A Timedistributed Version Difference of Structural Similarity (DSSIM loss function)

DSSIM_TD_loss.py

from keras.objectives import *
import keras_contrib.backend as KC

class DSSIM_TD():
    def __init__(self, k1=0.01, k2=0.03, kernel_size=3, max_value=1.0,n_imgs=0):
        """
        A Timedistributed Version Difference of Structural Similarity (DSSIM loss function). Clipped between 0 and 0.5
        Note : You should add a regularization term like a l2 loss in addition to this one.
        Note : In theano, the `kernel_size` must be a factor of the output size. So 3 could
               not be the `kernel_size` for an output of 32.
        # Arguments
            k1: Parameter of the SSIM (default 0.01)
            k2: Parameter of the SSIM (default 0.03)
            kernel_size: Size of the sliding window (default 3)
            max_value: Max value of the output (default 1.0)
            n_imgs: Number of prediected Images in each sample
        """
        self.__name__ = 'DSSIMObjective_TD'
        self.kernel_size = kernel_size
        self.k1 = k1
        self.k2 = k2
        self.max_value = max_value
        self.c1 = (self.k1 * self.max_value) ** 2
        self.c2 = (self.k2 * self.max_value) ** 2
        self.dim_ordering = K.image_data_format()
        self.backend = KC.backend()
        self.n_imgs=n_imgs
        if n_imgs==0:
            raise ValueError('n_imgs can\'t equal zero,use DSSIMObjective instead')

    def __int_shape(self, x):
        return KC.int_shape(x) if self.backend == 'tensorflow' else KC.shape(x)

    def __call__(self, y_true_o, y_pred_o):
      
        # There are additional parameters for this function
        # Note: some of the 'modes' for edge behavior do not yet have a gradient definition in the Theano tree
        #   and cannot be used for learning
        
        kernel = [self.kernel_size, self.kernel_size]
        ks=[]
        
        total_t=self.max_value*self.n_imgs
        
        for in_image in range(0,self.n_imgs):
           y_true=y_true_o[:,in_image]
           y_pred=y_pred_o[:,in_image]
            
           y_true = KC.reshape(y_true, [-1] + list(self.__int_shape(y_pred)[1:]))
           y_pred = KC.reshape(y_pred, [-1] + list(self.__int_shape(y_pred)[1:]))
 
   
           patches_pred = KC.extract_image_patches(y_pred, kernel, kernel, 'valid', self.dim_ordering)
           patches_true = KC.extract_image_patches(y_true, kernel, kernel, 'valid', self.dim_ordering)

           # Reshape to get the var in the cells
           bs, w, h, c1, c2, c3 = self.__int_shape(patches_pred)
           patches_pred = KC.reshape(patches_pred, [-1, w, h, c1 * c2 * c3])
           patches_true = KC.reshape(patches_true, [-1, w, h, c1 * c2 * c3])
           # Get mean
           u_true = KC.mean(patches_true, axis=-1)
           u_pred = KC.mean(patches_pred, axis=-1)
           # Get variance
           var_true = K.var(patches_true, axis=-1)
           var_pred = K.var(patches_pred, axis=-1)
           # Get std dev
           covar_true_pred = K.mean(patches_true * patches_pred, axis=-1) - u_true * u_pred

           ssim = (2 * u_true * u_pred + self.c1) * (2 * covar_true_pred + self.c2)
           denom = (K.square(u_true) + K.square(u_pred) + self.c1) * (var_pred + var_true + self.c2)
           ssim /= denom  # no need for clipping, c1 and c2 make the denom non-zero
           ks.append(K.mean((1.0 - ssim) / 2.0))
        all_summed=0
        for i in ks:
            all_summed+=i
        return all_summed/total_t