bcmi / CaGNet-Zero-Shot-Semantic-Segmentation

Code for our ACMMM2020 paper "Context-aware Feature Generation for Zero-shot Semantic Segmentation".
233 stars 14 forks source link

Questions about Finetuning part #17

Closed Ferenas closed 1 year ago

Ferenas commented 2 years ago

Hi, thanks for this excellent work! However, I have one problem with the settings in Finetuning part. In this paper, you mention "...The corresponding label map is Y(s/m) with pixel-wise label vector...", which contains both the unseen and seen part. Does it mean the ground-truth label map of the unseen class is involved during the finetuning process? Otherwise how to generate the label map of the unseen class during the fine-tuning process?