Soft Dice Loss

Problem

Compute the soft Dice loss between a probability map and a binary target.

Return 1 - (2 * sum(p * t) + eps) / (sum(p) + sum(t) + eps), rounded to 4 decimal places. eps guards against division by zero.

case 1

in: [[[0.8,0.2]],[[1,0]],0]

expected: 0.2

case 2

in: [[[1,0]],[[1,0]],0.000001]

expected: 0

case 3

in: [[[0.5]],[[1]],0.000001]

expected: 0.3333

case 4

in: [[[0.9,0.1],[0.1,0.9]],[[1,0],[0,1]],0]

expected: 0.1

Starter codePython

def soft_dice_loss(prob, target, eps):
    # Your code here
    pass

4 test cases · run your code against them in the editor