Direct Entropy Minimization

• On the source domain we train our model,
• as usual using a supervised loss
• For the target domain, we do not have annotations and we can no longer use the segmentation loss to train
• supervision signal that could leverage visual information from the target samples, in spite of the lack of annotations
• constrain
• to produce high-confident predictions on target samples similarly to source samples
• Entropy loss ${\mathcal{L}}_{ent}$ to maximize directly the prediction confidence in the target domain.
• Shannon Entropy