Hi guys,
I am implementing the deeplabv3+, imitating the pattern of deeplabv3.py,
but I don't quite understand the meaning for DeepLabHead,
so do I need to put the upsampling operations in the DeepLabHead?
Any answer and idea will be appreciated!
Hi,
In current implementation, the upsampling is done outside of DeepLabHead, see https://github.com/pytorch/vision/blob/819b2e9f39648711ab06562fc8f03ad001cf10be/torchvision/models/segmentation/_utils.py#L25
Let us know if you have further questions!
@fmassa , hi, Fancisco.
Thanks sincerely for your guides!