If I were to restart a semantic segmentation model project again, I would choose the loss function by considering the following.
Is there any imbalance in your data? ⇒（NO）⇒ Binary Cross Entropy
⇓
(YES)
⇓
Is the area you want detect relatively small? ⇒（NO）⇒ Dice Loss
⇓
(YES)
⇓
Focal Tversky Loss
