How you train on both Cityscapes and GTA5 datasets?

wasidennis / AdaptSegNet

Learning to Adapt Structured Output Space for Semantic Segmentation, CVPR 2018 (spotlight)

847 stars 203 forks source link

How you train on both Cityscapes and GTA5 datasets? #63

Closed Ha0Tang closed 5 years ago

Ha0Tang commented 5 years ago

Cityscapes dataset has only 2975 images, while the GTA5 dataset has 24966 images, I wonder how you load data and train the segmentation network on both datasets.