Custom loss function in Keras with TensorFlow Backend for images

Question

I am new to neural networks and keras and am having trouble writing this custom loss function:

I am using TensorFlow as backend. I saw other examples and wrote the loss function in this way:

from keras import backend as K
def depth_loss_func(pred_depth,actual_depth):
    n = pred_depth.shape[0]
    di = K.log(pred_depth)-K.log(actual_depth)
    di_sq = K.square(di)
    sum_d = K.sum(di)
    sum_d_sq = K.sum(di_sq)
    loss = ((1/n)*sum_d_sq)-((1/(n*n))*sum_d*sum_d) # getting an error in this step
    return loss

The error I am getting is : TypeError: unsupported operand type(s) for /: 'int' and 'Dimension'

Also I am not sure how to incorporate the learning rate in the loss function. Thanks for your help.

Answer 1

Instead of using "n", which seems not to be the most elegant way in my opinion, try using the K.mean function:

di = K.log(pred_depth)-K.log(actual_depth)

di_mean = K.mean(di)
sq_mean = K.mean(K.square(di))

loss = (sq_mean - (lamb*di_mean*di_mean)) # getting an error in this step