PASCAL-Context dataset

[zhaojiancheng007]

Hi ! Thanks for your interesting work, I got a question on the dataset you use:
In the paper and configuration in the codebase, you mentioned you use 'PASCAL-Context' dataset, which cntains 61 semantic segmentation classes, but your model predict 21 classes, which is the congiguration of the 'PASCAL VOC' dataset. I got a little confused, can you teach me more details about your dataset configurations?