Deep neural networks are susceptible to learn biased models with entangled feature
representations, which may lead to subpar performances on various downstream tasks. This
is particularly true for under-represented classes, where a lack of diversity in the data
exacerbates the tendency. This limitation has been addressed mostly in classification tasks,
but there is little study on additional challenges that may appear in more complex dense
prediction problems including semantic segmentation. To this end, we propose a model …

To verify whether the proposed method successfully leads disentangled representation
learning or not, we conduct a simple experiment to check reconstructability of a dropped
class. Through this experiment, our goal is to confirm a following hypothesis: If the model
has learned disentangled representations, the gradients would truly propagate class-
specific information and enable class drop in the feature representation. In other words,
there would be less redundancy between feature representations among different classes …