Architecture of deeplabv3_resnet50

How does the architecture of deeplabv3_resnet50 look like?

I am using this command to get the model:
model = torchvision.models.segmentation.deeplabv3_resnet50 (pretrained=False , progress=True , num_classes=21 , aux_loss=None , **kwargs )

I can then get some impression of how the model looks like just by printing model, but this doesn’t give me a good intuitive way of visualizing the model.

Printing out the model wouldn’t show the computation graph and would only print the child modules, so I agree that this would not be sufficient to “see” the structure.
You could check out e.g. PyTorchViz to visualize the computation graph in case that’s helpful.

PS: Often I also take a look at the source code, but for segmentation/detection models this is unfortunately also not trivial.

1 Like

Thanks @ptrblck ! PyTorchViz was exactly what I was looking for :grinning:

1 Like