Write Dice loss and its role in segmentation.

medium

Answer

$( \mid P \mid \; + \; \mid G \mid )$
$Directly optimizes overlap between predicted and ground-truth masks.$
$Robust to foreground/background imbalance in medical images where the target is 1% of pixels.$
$Often combined with BCE (Dice + BCE, or Dice + focal) for stability and calibration.$

Check yourself — multiple choice

Same as cross-entropy
1 - 2·|P∩G|/ $( \mid P \mid + \mid G \mid )$ — direct overlap loss, robust to foreground/background imbalance in segmentation
Only for regression
Ignores intersection

Dice loss: overlap-based, robust to imbalance — standard in medical segmentation.

#losses#segmentation#computer-vision

214 interview questions in this topic.