I wanted to use the coco 2017 stuff dataset for training a semantic segmentation network,
but I found that in the png segmentation masks that coco provided seems to have 3 channels.
I would very much appreciate it if you could explain how to obtain the 1 channel segmentation mask.
I wanted to use the coco 2017 stuff dataset for training a semantic segmentation network, but I found that in the png segmentation masks that coco provided seems to have 3 channels. I would very much appreciate it if you could explain how to obtain the 1 channel segmentation mask.
Thanks